?
Naslov in reference |
Title and reference |
POROČILO KOMISIJE EVROPSKEMU PARLAMENTU IN SVETU o izvajanju Uredbe (EU) 2021/1232 Evropskega parlamenta in Sveta z dne 14. julija 2021 o začasnem odstopanju od nekaterih določb Direktive 2002/58/ES glede uporabe tehnologij s strani ponudnikov medosebnih komunikacijskih storitev, neodvisnih od številke, za obdelavo osebnih in drugih podatkov za namene boja proti spolni zlorabi otrok na spletu POROČILO KOMISIJE EVROPSKEMU PARLAMENTU IN SVETU o izvajanju Uredbe (EU) 2021/1232 Evropskega parlamenta in Sveta z dne 14. julija 2021 o začasnem odstopanju od nekaterih določb Direktive 2002/58/ES glede uporabe tehnologij s strani ponudnikov medosebnih komunikacijskih storitev, neodvisnih od številke, za obdelavo osebnih in drugih podatkov za namene boja proti spolni zlorabi otrok na spletu COM/2023/797 final |
REPORT FROM THE COMMISSION TO THE EUROPEAN PARLIAMENT AND THE COUNCIL on the implementation of Regulation (EU) 2021/1232 of the European Parliament and of the Council of 14 July 2021 on a temporary derogation from certain provisions of Directive 2002/58/EC as regards the use of technologies by providers of number-independent interpersonal communications services for the processing of personal and other data for the purpose of combating online child sexual abuse REPORT FROM THE COMMISSION TO THE EUROPEAN PARLIAMENT AND THE COUNCIL on the implementation of Regulation (EU) 2021/1232 of the European Parliament and of the Council of 14 July 2021 on a temporary derogation from certain provisions of Directive 2002/58/EC as regards the use of technologies by providers of number-independent interpersonal communications services for the processing of personal and other data for the purpose of combating online child sexual abuse COM/2023/797 final |
Datumi |
Dates |
|
|
Drugi podatki |
Miscellaneous information |
|
|
Klasifikacije |
Classifications |
|
|
Besedilo |
Text |
EVROPSKA KOMISIJA | EUROPEAN COMMISSION |
Bruselj, 19.12.2023 | Brussels, 19.12.2023 |
COM(2023) 797 final | COM(2023) 797 final |
POROČILO KOMISIJE EVROPSKEMU PARLAMENTU IN SVETU | REPORT FROM THE COMMISSION TO THE EUROPEAN PARLIAMENT AND THE COUNCIL |
o izvajanju Uredbe (EU) 2021/1232 Evropskega parlamenta in Sveta z dne 14. julija 2021 o začasnem odstopanju od nekaterih določb Direktive 2002/58/ES glede uporabe tehnologij s strani ponudnikov medosebnih komunikacijskih storitev, neodvisnih od številke, za obdelavo osebnih in drugih podatkov za namene boja proti spolni zlorabi otrok na spletu | on the implementation of Regulation (EU) 2021/1232 of the European Parliament and of the Council of 14 July 2021 on a temporary derogation from certain provisions of Directive 2002/58/EC as regards the use of technologies by providers of number-independent interpersonal communications services for the processing of personal and other data for the purpose of combating online child sexual abuse |
KAZALO | CONTENTS |
SEZNAM POJMOV IN KRATIC | LIST OF TERMS AND ACRONYMS |
1. UVOD | 1. INTRODUCTION |
2. IZVEDBENI UKREPI | 2. IMPLEMENTATION MEASURES |
2.1 Obdelava osebnih podatkov s strani ponudnikov (člen 3(g)(vii)) | 2.1. Processing of personal data by providers (Article 3 (g) (vii)) |
2.1.1 Vrsta in količine obdelanih podatkov | 2.1.1. Type and volumes of data processed |
2.1.2 Razlogi za obdelavo na podlagi Uredbe (EU) 2016/679 | 2.1.2. Grounds for processing pursuant to Regulation (EU) 2016/679 |
2.1.3 Razlogi za prenose osebnih podatkov iz Unije v skladu s Poglavjem V GDPR, kadar je to ustrezno | 2.1.3. Ground for transfers of personal data outside the Union pursuant to Chapter V of the GDPR, where applicable |
2.1.4 Število ugotovljenih primerov spolne zlorabe otrok na spletu, pri čemer se razlikuje med posnetki spolne zlorabe otrok na spletu in pridobivanjem otrok za spolne namene | 2.1.4. Number of cases of online child sexual abuse identified, differentiating between CSAM and solicitation of children |
2.1.5 Pravna sredstva uporabnikov in izid | 2.1.5. User redress and outcome |
2.1.6 Število in deleži napak (lažno pozitivnih primerov) pri različnih uporabljenih tehnologijah | 2.1.6. Number and ratios of errors (false positives) of the different technologies used |
2.1.7 Ukrepi, sprejeti za omejitev stopnje napak, in dosežena stopnja napak | 2.1.7. Measures applied to limit the error rate and the error rate achieved |
2.1.8 Politika hrambe podatkov in ukrepi, sprejeti za zaščito podatkov na podlagi splošne uredbe o varstvu podatkov | 2.1.8. The retention policy and the data protection safeguards applied pursuant to the GDPR |
2.1.9 Imena organizacij, ki delujejo v javnem interesu proti spolni zlorabi otrok, katerih podatki so bili posredovani na podlagi Uredbe | 2.1.9. The names of the organisations acting in the public interest against child sexual abuse with which data has been shared pursuant to this Regulation |
2.2 Statistični podatki Komisije (člen 8) | 2.2. Member States' statistics (Article 8) |
2.2.1 Skupno število prijav odkritih spolnih zlorab otrok na spletu | 2.2.1. The total number of reports of detected online child sexual abuse |
2.2.2 Število identificiranih otrok | 2.2.2. The number of children identified |
2.2.3 Število obsojenih storilcev | 2.2.3. The number of perpetrators convicted |
2.3 Razvoj tehnološkega napredka | 2.3. Developments in technological progress |
2.3.1 Odkrivanje znanih posnetkov spolne zlorabe otrok | 2.3.1. Known CSAM detection |
2.3.2 Odkrivanje novih posnetkov spolne zlorabe otrok | 2.3.2. New CSAM detection |
2.3.3 Odkrivanje pridobivanja otrok za spolne namene | 2.3.3. Grooming detection |
2.3.4 Novi izzivi, ki jih prinašajo klepetalni boti z umetno inteligenco in generatorji umetnin/podob | 2.3.4. Artificial intelligence chatbots and art/image generators |
3. SKLEPNE UGOTOVITVE | 3. CONCLUSIONS |
SEZNAM POJMOV IN KRATIC | LIST OF TERMS AND ACRONYMS |
Pojem ali kratica | Opredelitev | Term/Acronym | Definition |
UI | Umetna inteligenca | AI | Artificial Intelligence |
API | Vmesniki za aplikacijsko programiranje | API | Application Programming Interfaces |
CG-CSAM | Računalniško ustvarjeni posnetki spolne zlorabe otrok | CG-CSAM | Computer-generated Child Sexual Abuse Material |
ChatGPT | ChatGPT (generativni prednaučeni transformer za klepet) je oblika generativne umetne inteligence. To je velik jezikovni na modelih temelječ klepetalni bot, ki ga je razvila organizacija OpenAI in ki uporabnikom omogoča, da izpopolnijo in usmerijo pogovor glede na želeno dolžino, obliko, slog, raven podrobnosti in jezik. | ChatGPT | ChatGPT (Chat Generative Pre-trained Transformer) is a form or generative AI. It is a large language model-based chatbot, developed by OpenAI, that enables users to refine and steer a conversation towards a desired length, format, style, level of detail, and language. |
Klasifikatorji | Oblika umetne inteligence, algoritem, ki podatke razvršča v označene razrede ali kategorije | Classifiers | A form of artificial intelligence, an algorithm that sorts data into labelled classes or categories |
Klasifikator Content Safety API | Googlov klasifikator Content Safety API s programskim dostopom in umetno inteligenco pomaga pri klasifikaciji in prednostnem razvrščanju milijarde podob za pregled | Content Safety API classifier | Google’s Content Safety API classifier uses programmatic access and artificial intelligence to help classify and prioritise billions of images for review |
SZO | Spolna zloraba otrok | CSA | Child Sexual Abuse |
CSAI Match | CSAI Match je tehnologija, ki so jo razvili inženirji podjetja YouTube za prepoznavanje ponovnih prenosov videoposnetkov, pri katerih je bilo že predhodno ugotovljeno, da gre za spolno zlorabo otrok | CSAI Match | CSAI Match is a technology developed by YouTube engineers to identify re-uploads of previously identified child sexual abuse in videos |
CSAM | Posnetki spolne zlorabe otrok, npr. podobe in videoposnetki, ki prikazujejo spolno zlorabo otrok | CSAM | Child Sexual Abuse Material, e.g. images and videos depicting CSA |
Direktiva o boju proti spolni zlorabi otrok | Direktiva 2011/93/EU Evropskega parlamenta in Sveta z dne 13. decembra 2011 o boju proti spolni zlorabi in spolnemu izkoriščanju otrok ter otroški pornografiji in nadomestitvi Okvirnega sklepa Sveta 2004/68/PNZ, UL L 335, 17.12.2011. str. 1 | CSA Directive | Directive 2011/93/EU of the European Parliament and of the Council of 13 December 2011 on combating the sexual abuse and sexual exploitation of children and child pornography, and replacing Council Framework Decision 2004/68/JHA, OJ L 335, 17.12.2011, p. 1–14 |
CSEA | Spolno izkoriščanje in zloraba otrok | CSEA | Child Sexual Exploitation and Abuse |
Spolna zloraba otrok na spletu | Uveljavljen pojem, ki se uporablja za tri vrste spolne zlorabe otrok, opredeljene v direktivi o boju proti spolni zlorabi otrok, in sicer: otroško pornografijo, pornografsko predstavo in pridobivanje otrok za spolne namene, kot je opredeljen v začasni uredbi (člen 2(4)) | CSA online | The common term used for the three types of child sexual abuse, defined in the CSA Directive, namely: child pornography, pornographic performance and solicitation of children (‘grooming’), as defined in the Interim Regulation (Article 2(4)) |
EU | Evropska unija | EU | European Union |
GDPR | Uredba (EU) 2016/679 Evropskega parlamenta in Sveta z dne 27. aprila 2016 o varstvu posameznikov pri obdelavi osebnih podatkov in o prostem pretoku takih podatkov ter o razveljavitvi Direktive 95/46/ES (Splošna uredba o varstvu podatkov), UL L 119, 4.5.2016, str. 1 | GDPR | Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation), OJ L 119, 4.5.2016, p. 1–88 |
Pridobivanje otrok za spolne namene | Storilci kaznivih dejanj krepijo zaupanje in odnos z otrokom, da bi pridobili dostop do mladoletnika za namene spolnega izkoriščanja ali zlorabe. Pridobivanje otrok za spolne namene je opredeljeno v členu 6 direktive o boju proti spolni zlorabi otrok | Grooming | Offenders building trust and a relationship with a child in an effort to gain access to the minor for sexual exploitation or abuse. Formally known as solicitation of children, as defined in Article 6 of the CSA Directive |
Edinstven nerekonvertibilen digitalni podpis | Edinstvena digitalna koda, ustvarjena z matematičnim algoritmom („zgoščevanje“), ki postane podpis zadevne datoteke ali njegova zgoščena vrednost | Hash | A unique digital code created by a mathematical algorithm (“hashing”) that becomes this file’s signature, or its hash value |
Začasna uredba | Uredba (EU) 2021/1232 Evropskega parlamenta in Sveta z dne 14. julija 2021 o začasnem odstopanju od nekaterih določb Direktive 2002/58/ES glede uporabe tehnologij s strani ponudnikov medosebnih komunikacijskih storitev, neodvisnih od številke, za obdelavo osebnih in drugih podatkov za namene boja proti spolni zlorabi otrok na spletu, UL L 274, 30.7.2021, str. 41 | Interim Regulation | Regulation (EU) 2021/1232 of the European Parliament and of the Council of 14 July 2021 on a temporary derogation from certain provisions of Directive 2002/58/EC as regards the use of technologies by providers of number-independent interpersonal communications services for the processing of personal and other data for the purpose of combating online child sexual abuse, OJ L 274, 30.7.2021, p. 41–51 |
LLM | Velik jezikovni model (LLM) je vrsta modela umetne inteligence, ki je bil z algoritmi globokega učenja naučen za prepoznavanje, ustvarjanje, prevajanje in/ali povzemanje velikih količin pisnega človeškega jezika in besedilnih podatkov. | LLM | A large language model (LLM) is a type of artificial intelligence model that has been trained through deep learning algorithms to recognise, generate, translate, and/or summarise vast quantities of written human language and textual data. |
MD5 | Kriptografski algoritem za kodo za ugotavljanje avtentičnosti sporočila za uporabo na internetu | MD5 | Cryptographic message authentication code algorithm for use on the internet |
Meta SSN++ | Model umetne inteligence, ki ga je razvilo podjetje Meta, ki lahko zazna skoraj popolne dvojnike | Meta SSN++ | An AI model, developed by Meta, that can detect near-exact duplicates. |
Center NCMEC | Nacionalni center za pogrešane in izkoriščane otroke (zasebna nepridobitna organizacija v ZDA), ki mu morajo ponudniki spletnih storitev v skladu s pravom ZDA prijaviti primere morebitne spolne zlorabe otrok, ki jih odkrijejo v svojih omrežjih | NCMEC | National Centre for Missing and Exploited Children (US private, non-profit organisation) to which online service providers are required to report under US law instances of potential child sexual abuse that they find in their networks |
PDQ in TMK+PDQF | Orodji, ki ju Facebook uporablja za odkrivanje škodljive vsebine. PDQ je tehnologija za iskanje ujemajočih se fotografij; TMK+PDQF je tehnologija za iskanje ujemajočih se videoposnetkov. | PDQ and TMK+PDQF | Tools used by Facebook to detect harmful content. PDQ is a photo matching technology; TMK+PDQF is a video-matching technology. |
PhotoDNA | Najpogosteje uporabljeno orodje na podlagi tehnologije zgoščevanja, ki je na voljo brezplačno in temelji na licenčni pogodbi, ki je prilagojena tako, da se preprečijo zlorabe in da se lahko uporablja izključno za odkrivanje spolne zlorabe otrok | PhotoDNA | The most widely used tool based on hashing technology, available free of charge, based on a licensing agreement tailored to avoid abuse and use for any other purpose than the detection of CSA |
1.UVOD | 1.INTRODUCTION |
Komisija mora v skladu s členom 9 začasne uredbe (v nadaljnjem besedilu tudi: Uredba) pripraviti poročilo o izvajanju Uredbe, in sicer na podlagi poročil, ki jih predložijo ponudniki medosebnih komunikacijskih storitev (v nadaljnjem besedilu: ponudniki) o obdelavi osebnih podatkov in o statističnih podatkih, ki jih predložijo države članice. V skladu z navedeno določbo Komisija v poročilu o izvajanju upošteva zlasti naslednje: | Article 9 of the Interim Regulation (hereafter also: “the Regulation”) obliges the Commission to prepare a report on its implementation, based on reports submitted by providers of interpersonal communications services (hereafter: “providers”) on the processing of personal data and on the statistics provided by Member States. According to the aforementioned provision, in the implementation report, the Commission shall consider, in particular, |
(a) ustrezne pogoje za obdelavo zadevnih osebnih podatkov in drugih podatkov, ki se obdelujejo na podlagi Uredbe; | (a) the relevant conditions for the processing of relevant personal data and other data processed under the Regulation; |
(b) sorazmernost odstopanja, ki ga določa Uredba, vključno z oceno statističnih podatkov, ki jih države članice predložijo na podlagi člena 8 Uredbe; | (b) the proportionality of the derogation provided for by the Regulation, including an assessment of the statistics submitted by the Member States pursuant to its Article 8; |
(c) razvoj na tehnološkem področju, ki zadeva dejavnosti iz Uredbe, in v kolikšni meri se z njim izboljša natančnost ter zmanjšajo število in deleži napak (lažno pozitivnih primerov). | (c) developments in technological progress regarding the activities covered by the Regulation and the extent to which such developments improve accuracy and reduce the numbers and ratios of errors (false positives). |
To poročilo o izvajanju na podlagi začasne uredbe temelji na podatkih, pridobljenih s poročili ponudnikov in držav članic v skladu z njenim členom 3(1)(g)(vii) oziroma členom 8. Taka poročila so razkrila velike razlike, kar zadeva razpoložljivost podatkov in vrste zbranih podatkov ter s tem tudi primerljivost podatkov, ki jih zberejo ponudniki in države članice. Ker Uredba ne zagotavlja predloge za poročanje, so ponudniki delili različne vrste informacij, ki niso bile nujno primerljive; službe Komisije so zato izvedle nadaljnje ukrepe, da bi zagotovile pravilno razlago podatkov. Večina držav članic ni mogla pravočasno predložiti podatkov, veliko držav članic pa ni moglo predložiti nobenih podatkov do objave tega poročila. To je pomembno vplivalo na časovnico, popolnost in koristnost poročila. Kljub prizadevanjem za zagotovitev skladnosti in primerljivosti podatkov še vedno obstajajo velike razlike, ki so prikazane v preglednicah v nadaljevanju, ki pa prav tako ne vsebujejo podatkov za vse ponudnike ali države članice o vseh vidikih. | This implementation report under the Interim Regulation builds on the data obtained through reporting by providers and Member States pursuant to its Article 3(1)(g)(vii) and Article 8, respectively. Such reporting brought to light significant disparities in the availability of data, the types of data collected, and thus also the comparability of the data collected by providers and Member States. As the Regulation does not provide a template for the reporting, providers shared different types of information which were not necessarily comparable; the Commission services therefore engaged in follow-up to ensure the correct interpretation of the data. The majority of Member States were unable to provide data in time and a number of them have been unable to provide any data until the publication of this report. This had a significant impact on the timing, completeness and usefulness of the report. Despite the efforts to ensure coherence of the data and comparability, significant disparities remain which are reflected in the below tables, which also do not contain data for all providers or Member States on all points. |
Namen tega poročila o izvajanju je zagotoviti dejanski pregled stanja v zvezi z izvajanjem začasne uredbe, na podlagi razpoložljivih podatkov. Poročilo ne vsebuje razlag Uredbe in se ne opredeljuje glede načina tolmačenja in uporabe Uredbe v praksi. | This implementation report seeks to give a factual overview of the state of play in connection to the implementation of the Interim Regulation, based on the available data. The report does not contain any interpretations of the Regulation and does not take any position on the manner in which it has been interpreted and applied in practice. |
2.IZVEDBENI UKREPI | 2.IMPLEMENTATION MEASURES |
2.1Obdelava osebnih podatkov s strani ponudnikov (člen 3(g)(vii)) | 2.1.Processing of personal data by providers (Article 3 (g) (vii)) |
Člen 3(g)(vii) začasne uredbe določa pogoje, pod katerimi ponudniki, ki delujejo na podlagi odstopanja, določenega v navedenem členu, do 3. februarja 2022 in nato vsako leto do 31. januarja objavijo ter pristojnemu nadzornemu organu in Komisiji predložijo poročilo o obdelavi osebnih podatkov na podlagi Uredbe. Podjetja Google, LinkedIn, Meta, Microsoft in X (prej Twitter) 1 so predložila poročila za leti 2021 in 2022. | Article 3(g)(vii) of the Interim Regulation lays down the conditions for providers acting under the derogation contained therein to publish and submit to the competent supervisory authority and to the Commission, by 3 February 2022, and by 31 January every year thereafter, a report on the processing of personal data under this Regulation. Google, LinkedIn, Meta, Microsoft, and X (former Twitter) 1 submitted reports for 2021 and 2022. |
2.1.1.Vrsta in količine obdelanih podatkov | 2.1.1.Type and volumes of data processed |
Ponudniki so poročali o obdelavi podatkov o vsebini in prometu. | Providers reported processing both content and traffic data. |
V zvezi s podatki o vsebini, obdelanimi za odkrivanje spolne zlorabe otrok na spletu, so vsi navedeni ponudniki navedli podobe in videoposnetke. Večinoma so za odkrivanje ujemanj s posnetki, pri katerih je bilo predhodno ugotovljeno, da gre za spolno zlorabo otrok, uporabljali tehnologiji za iskanje ujemanj edinstvenega, nerekonvertibilnega digitalnega podpisa PhotoDNA in MD5. Googlovo orodje CSAI Match se je uporabljalo za ustvarjanje digitalnih prstnih odtisov videoposnetkov na platformah in njihovo primerjavo z datotekami v Googlovem/YouTubovem repozitoriju prstnih odtisov (LinkedIn). Na primer, podjetje Google je poročalo tudi o uporabi avtomatizirane tehnologije (strojno učenje z umetno inteligenco) in človeškem pregledu. Podjetji Google in LinkedIn sta potrdili, da sta prepoznali tudi posnetke spolne zlorabe otrok, ki se niso ujemali s posnetki, pri katerih je bilo predhodno ugotovljeno, da gre za spolno zlorabo otrok. Noben od petih ponudnikov, ki so predložili podatke, ni sporočil podatkov, da je odkril pridobivanje otrok za spolne namene s pomočjo odkrivanja besedil v okviru odstopanja, ki ga zagotavlja Uredba. | As regards content data processed to detect online child sexual abuse, all aforementioned providers mentioned images and videos. Mostly they relied on the hash matching technologies PhotoDNA and MD5 to detect matches of previously identified child sexual abuse material (hereafter: ”CSAM”). Google’s CSAI Match tool was used to create digital fingerprints of videos on platforms and compare them with the files in Google/YouTube’s fingerprint repository (LinkedIn). The use of automated technology (artificial intelligence machine learning) and human review was equally reported (e.g. by Google). Google and LinkedIn confirmed identifying also CSAM that did not match previously identified CSAM. None of the five providers that submitted data reported data on detecting solicitation of children via text detection under the scope of the derogation provided by this Regulation. |
V zvezi z zbranimi podatki o prometu ter ustreznimi količinami različnih vrst obdelanih podatkov o vsebini in prometu so se poročila ponudnikov precej razlikovala. | As for traffic data collected and the respective volumes of the different types of content and traffic data processed, the reports of providers varied substantially. |
Podatki o prometu, ki so jih zbrali ponudniki in jih vključili v poročila v sistemu CyberTipline za ameriški Nacionalni center za pogrešane in izkoriščane otroke („NCMEC“) vključujejo naslednje podatke (vse ali izbor podatkov): | Traffic data collected by providers and included in CyberTipline reports to the U.S. National Center for Missing and Exploited Children (hereafter: “NCMEC”) include (all or a selection) of the following data: |
a)podatke, povezane z uporabnikom/poročevalcem/računom (Google, LinkedIn, Microsoft, X); | a)User/reportee/account related data (Google, LinkedIn, Microsoft, X); |
b)metapodatke, povezane s podatki o vsebini/transakciji (Google, LinkedIn, Microsoft); | b)Metadata related to content/transactional data (Google, LinkedIn, Microsoft); |
c)podatke, povezane z morebitno žrtvijo (Google); | c)Data related to a potential victim (Google); |
d)podatke o dejanjih zlorabe (Google). | d)Abuse operations data (Google). |
Kar zadeva količine podatkov, obdelanih na podlagi začasne uredbe, je podjetje LinkedIn poročalo o obdelavi 8 milijonov podob in videoposnetkov, ki so izvirali iz EU, med 14. julijem in 31. decembrom 2021, ter o 21 milijonih podob in 63 000 videoposnetkih iz EU leta 2022. Podjetje Microsoft je poročalo o obdelavi, za namene Uredbe, 8,9 milijarde podob in videoposnetkov po vsem svetu med julijem in decembrom 2021 ter o 12,3 milijarde vsebinah po vsem svetu leta 2022; podatki za EU niso bili na voljo, zato ni mogoče oblikovati sklepov za namene tega poročila. Drugi ponudniki niso zagotovili informacij o količinah obdelanih podatkov. Od petih ponudnikov, ki so poročali, je torej samo eden predložil podatke na zahtevani ravni podrobnosti. | In terms of volumes of data processed under the Interim Regulation, LinkedIn reported processing 8 million images and videos originating from the EU between 14 July and 31 December 2021, and 21 million images and 63 000 videos originating from the EU in 2022. Microsoft reported processing for the purposes of the Regulation 8.9 billion images and videos globally between July and December 2021, and 12.3 billion content items globally in 2022; EU figures were not available, hence it is not possible to draw any conclusions for the purposes of this report. The other providers did not provide information on volumes of data processed. Of the five reporting providers, therefore, only one provided data at the requisite level of granularity. |
Da bi ponazoril splošni okvir, je center NCMEC poročal, da je leta 2022 prejel skupaj 87,2 milijona podob in videoposnetkov z vsega sveta ter 5,1 milijona podob in videoposnetkov iz EU, leta 2021 pa 84,8 milijona podob in videoposnetkov z vsega sveta ter 1,8 milijona podob in videoposnetkov iz EU. To je le gradivo, ki ga je ponudnik opredelil kot možne posnetke spolne zlorabe otrok, zato tega ni mogoče upoštevati kot kazalnik skupne količine podatkov, obdelanih na podlagi začasne uredbe. | To illustrate the overall context, NCMEC reported having received a total of 87.2 million images and videos globally and 5.1 million images and videos concerning the EU in 2022, and 84.8 million images and videos globally and 1.8 million images and videos concerning the EU in 2021. These are only the materials that have been identified as potential CSAM by a provider and therefore cannot be taken as indicative of the overall volumes of data processed under the Interim Regulation. |
2.1.2 Razlogi za obdelavo na podlagi Uredbe (EU) 2016/679 | 2.1.2. Grounds for processing pursuant to Regulation (EU) 2016/679 |
Ponudniki so poročali, da so se opirali na naslednje posebne razloge na podlagi Uredbe (EU) 2016/679 (v nadaljnjem besedilu: GDPR): | The providers reported relying on the following specific grounds pursuant to Regulation (EU) 2016/679 (hereafter: “GDPR”): |
·člen 6(1)(d) GDPR, ki določa, da je obdelava potrebna za zaščito življenjskih interesov otrok in tistih, ki so žrtve spolne zlorabe otrok na spletu (Google, Meta, X 2 ); | ·Article 6(1)(d) of the GDPR, i.e. processing necessary in order to protect the vital interests of children and those who are the victims of online child sexual abuse (Google, Meta, X 2 ); |
·člen 6(1)(e) GDPR, ki določa, da je obdelava potrebna za opravljanje naloge v javnem interesu (LinkedIn, Microsoft, Meta, X 3 ); | ·Article 6(1)(e) of the GDPR, i.e. processing necessary for the performance of tasks carried out in the public interest (LinkedIn, Microsoft, Meta, X 3 ); |
·člen 6(1)(f) GDPR, ki določa, da je obdelava potrebna zaradi zakonitih interesov: | ·Article 6(1)(f) of the GDPR, i.e. processing necessary for the purposes of the legitimate interests of: |
I.ponudnika, da odkrije, prepreči ali drugače obravnava spolno zlorabo otrok na spletu pri svojih storitvah ter da zaščiti druge uporabnike, stranke, partnerje in javnost pred tovrstno nezakonito vsebino (Google, Meta); | I.the provider to detect, prevent or otherwise address online child sexual abuse on their services and to protect other users, customers, partners, and the public from this form of illegal content (Google, Meta); |
II.žrtev spolne zlorabe otrok in organizacije, ki ji ponudnik prijavi spolno zlorabo otrok na spletu (npr. center NCMEC), da bi odkril, preprečil in odstranil spolno zlorabo otrok na spletu iz svojih storitev (Google). | II.victims of child sexual abuse and the organisation to whom the provider reports online child sexual abuse (e.g. NCMEC) to detect, prevent and remove online child sexual abuse from their services (Google). |
2.1.3.Razlogi za prenose osebnih podatkov iz Unije v skladu s Poglavjem V GDPR, kadar je to ustrezno | 2.1.3.Ground for transfers of personal data outside the Union pursuant to Chapter V of the GDPR, where applicable |
Vsi ponudniki so poročali, da so uporabljali standardna pogodbena določila, ki jih je Komisija odobrila v skladu členom 46(2)(c) GDPR. Kar zadeva prenose osebnih podatkov centru NCMEC, je podjetje LinkedIn tudi poročalo, da je uporabljalo odstopanje, kolikor je bilo to potrebno, ki ga dovoljuje člen 49(1) GDPR. | All providers reported relying on standard contractual clauses approved by the Commission pursuant to Article 46(2)(c) of the GDPR. For transfers of personal data to NCMEC, LinkedIn also reported relying on a derogation, to the extent applicable, permitted under Article 49(1) GDPR. |
2.1.4.Število ugotovljenih primerov spolne zlorabe otrok na spletu, pri čemer se razlikuje med posnetki spolne zlorabe otrok na spletu in pridobivanjem otrok za spolne namene | 2.1.4.Number of cases of online child sexual abuse identified, differentiating between CSAM and solicitation of children |
Preglednica št. 1: Število primerov spolne zlorabe otrok na spletu, ugotovljenih leta 2021 | Table n.1: Number of cases of online child sexual abuse identified in 2021 |
Ponudnik | Število primerov posnetkov spolne zlorabe otrok, ugotovljenih leta 2021 | Opombe | Provider | Number of cases of CSAM identified in 2021 | Comments |
Google | 33 računov za orodje Google Chat | To se nanaša na število računov za orodje Google Chat za uporabnike v EU, pri katerih je bila v obdobju od 2. avgusta 2021 do 31. decembra 2021 z avtomatiziranimi tehnologijami ugotovljena spolna zloraba otrok na spletu. Podatki o številu odkritih vsebin niso bili predloženi. | Google | 33 Google Chat accounts | This refers to the number of Google Chat accounts for EU users where online child sexual abuse was identified through automated technologies in the period from 2 August 2021 to 31 December 2021. No data was provided on the number of content items identified. |
LinkedIn | 31 vsebin | 31 vsebin je bilo z ročnim pregledom potrjenih kot posnetki spolne zlorabe otrok in prijavljenih centru NCMEC; 6 je bilo znanih, preostalih 25 pa je bilo neznanih posnetkov spolne zlorabe otrok. | LinkedIn | 31 content items | 31 content items were confirmed via manual review as CSAM and reported to NCMEC; 6 were known CSAM and the remaining 25 were unknown CSAM. |
Meta | 340 000 računov | Število računov, pri katerih je bilo odkrito pošiljanje najmanj ene medijske vsebine, ki je pomenila posnetek spolne zlorabe otrok, v nizih sporočil, ki so vključevali uporabnika EU, in sicer med 8. novembrom in 31. decembrom 2021. | Meta | 340 000 accounts | Number of accounts detected as sending at least one piece of media constituting CSAM, in message threads which included an EU user between 8 November 2021 to 31 December 2021. |
Microsoft | 6 600 vsebin | 6 600 vsebin (posamezne podobe ali videoposnetki), ki so bile potrjene kot posnetki spolne zlorabe otrok in odkrite v Evropski uniji med več kot 20 000 vsebinami, odkritimi po svetu med julijem 2021 in decembrom 2021. | Microsoft | 6 600 content items | 6 600 content items (single image or video) confirmed as CSAM detected from the European Union out of over 20 000 content items identified globally between July 2021 to December 2021. |
X (prej Twitter) | 532 898 računov | Računi (ni jasno, ali samo v EU ali po vsem svetu), preklicani zaradi kršitve politike podjetja X glede preprečevanja spolnega izkoriščanja otrok med 2. avgustom 2021 in 31. decembrom 2021. | X (former Twitter) | 532 898 accounts | Accounts (unclear if only from the EU or globally) suspended for violating X’s child sexual exploitation policy between 2 August 2021 to 31 December 2021. |
V primeru podjetja X iz predloženih podatkov ni jasno, ali se nanašajo izključno na storitve, ki spadajo na področje uporabe začasne uredbe (medosebne komunikacijske storitve, neodvisne od številke), ali navedeno število zajema tudi druge storitve (na primer storitve informacijske družbe). To velja za vse podatke, povezane s podjetjem X v tem poročilu. | In the case of X, it is unclear from the data provided whether it relates exclusively to services falling into the scope of the Interim Regulation (number-independent interpersonal communications services) or whether that number also comprises other services (such as information society services). This concerns all the data related to X in this report. |
Preglednica št. 2: Število primerov spolne zlorabe otrok na spletu, ugotovljenih leta 2022 | Table n.2: Number of cases of online child sexual abuse identified in 2022 |
Ponudnik | Število primerov posnetkov spolne zlorabe otrok, ugotovljenih leta 2022 | Opombe | Provider | Number of cases of CSAM identified in 2022 | Comments |
Google | 2 045 vsebin | Vsebine, odkrite v 752 Googlovih računih z avtomatiziranimi tehnologijami za uporabnike v EU in prijavljene centru NCMEC. | Google | 2 045 content items | Content items identified and reported to NCMEC in 752 Google accounts through automated technologies for EU users. |
LinkedIn | 2 vsebini | LinkedIn je odkril dve podobi in nič videoposnetkov, potrjenih kot posnetki spolne zlorabe otrok. | LinkedIn | 2 content items | LinkedIn detected 2 images and 0 videos confirmed as CSAM. |
Meta | 6,6 milijona vsebin | Medijske vsebine, ki so pomenile posnetke spolne zlorabe otrok in zoper katere so bili uvedeni ukrepi ter ki so bile odkrite s tehnologijo podjetja Meta za iskanje ujemajočih se medijev, v nizih sporočil, | ki so vključevali uporabnika v EU. | Meta | 6.6 million content items | Pieces of media constituting CSAM actioned that were detected using Meta’s media matching technology, in message | threads which included an EU user. |
Microsoft | 12 800 vsebin | 12 800 vsebin (posamezne podobe ali videoposnetki), ki so bile potrjene kot posnetki spolne zlorabe otrok in odkrite v EU med več kot 50 000 vsebinami, odkritimi po svetu leta 2022. | Microsoft | 12 800 content items | 12 800 content items (single image or video) confirmed as CSAM detected from the EU out of over 50 000 content items identified globally in 2022. |
X (prej Twitter) | 2 348 712 računov | Računi (ni jasno, ali samo v EU ali po vsem svetu), preklicani zaradi kršitve politike podjetja X glede preprečevanja spolnega izkoriščanja otrok. | X (former Twitter) | 2 348 712 accounts | Accounts (unclear if only from the EU or globally) suspended for violating X’s child sexual exploitation policy. |
2.1.5.Pravna sredstva uporabnikov in izid | 2.1.5.User redress and outcome |
Ponudniki morajo v skladu s členom 3(1)(g)(iv) začasne uredbe uvesti ustrezne postopke in pritožbene mehanizme, da lahko uporabniki vložijo pritožbe. Poleg tega člen 5 Uredbe vsebuje pravila o pravnih sredstvih. | In accordance with Article 3(1)(g)(iv) of the Interim Regulation, providers are to establish appropriate procedures and redress mechanisms to ensure that users can lodge complaints with them. In addition, its Article 5 contains rules on judicial redress. |
Ponudniki so poročali, da so uvedli take notranje pritožbene postopke in mehanizme za uporabnike, katerih računi so bili omejeni zaradi deljenja posnetkov spolne zlorabe otrok in/ali katerih vsebina je bila odstranjena, ker je šlo za posnetke spolne zlorabe otrok, tako da se lahko pritožijo zoper omejitev/odstranitev in se prouči, ali so bile v njihovem primeru storjene napake. | Providers reported implementing such internal redress procedures and mechanisms for users whose accounts have been restricted for sharing CSAM and/or content removed as CSAM so that they can appeal their restriction/removal and have their case reviewed for errors. |
Poročali so o primerih, ko je uporabnik vložil pritožbo prek notranjega pritožbenega mehanizmu ali pri pravosodnem organu v zvezi z zadevami, ki spadajo na področje uporabe Uredbe v EU, ter o izidih takih pritožb. Noben ponudnik razen podjetja Microsoft (ki v letih 2021 in 2022 ni poročalo o pritožbah, vloženih na navedena načina) ni poročal o ločenih statističnih podatkih glede notranjega mehanizma za pritožbe in pravnih sredstev, zato so v spodnjih preglednicah zajeti notranji pritožbeni postopki in tudi postopki sodnega varstva. | They reported cases in which a user has lodged a complaint with the internal redress mechanism or with a judicial authority regarding matters in scope of the Regulation within the EU, and the outcomes of such complaints. No provider except Microsoft (which reported 0 complaints in both channels in 2021 and 2022) reported separate statistics on internal redress and judicial redress; as a result the tables below cover both internal and judicial redress procedures. |
Preglednica št. 3: Število primerov, v katerih je uporabnik vložil pritožbo prek notranjega mehanizma za pritožbe ali pri pravosodnem organu, ter izid takih pritožb v letu 2021 | Table n.3: Number of cases in which a user has lodged a complaint with the internal redress mechanism or with a judicial authority and the outcome of such complaints in 2021 |
Ponudnik | Število primerov pritožb uporabnikov | Število obnovljenih računov | Število ponovno objavljenih vsebin | Opombe | Provider | Number of cases of user complaints | Number of reinstated accounts | Number of reinstated contents | Comments |
Google | 8 | 0 | Ni na voljo. | Računi za orodje Google Chat, onemogočeni zaradi spolne zlorabe otrok na spletu, kjer se je uporabnik pritožil: 8. Noben račun ni bil obnovljen. | Google | 8 | 0 | n.a. | Google Chat accounts disabled for online child sexual abuse where the user appealed: 8. None was reinstated. |
LinkedIn | 0 | Ni na voljo. | Ni na voljo. | LinkedIn | 0 | n.a. | n.a. |
Meta | 4 900 | Ni na voljo. | 207 | 4 900 uporabnikov se je pritožilo. Po pritožbenem postopku je bila vsebina 207 uporabnikov obnovljena, ukrepi zoper račune pa razveljavljeni. | Meta | 4 900 | n.a. | 207 | 4 900 users appealed. Following the appeals process, 207 users had their content restored, and account actions reversed. |
Microsoft | 0 | Ni na voljo. | Ni na voljo. | Microsoft | 0 | n.a. | n.a. |
X (prej Twitter) | Pribl. 90 000 | Pribl. 3 000 | Ni na voljo. | Pribl. 90 000 pritožb. Podjetje X je obnovilo pribl. 3 000 od teh računov. | X (former Twitter) | ca 90 000 | ca 3 000 | n.a. | Ca 90 000 appeals. X reinstated ca 3 000 of those accounts. |
Preglednica št. 4: Število primerov, v katerih je uporabnik vložil pritožbo prek notranjega mehanizma za pritožbe ali pri pravosodnem organu, ter izid takih pritožb v letu 2022 | Table n.4: Number of cases in which a user has lodged a complaint with the internal redress mechanism or with a judicial authority and the outcome of such complaints in 2022. |
Ponudnik | Število primerov pritožb uporabnikov | Število obnovljenih računov | Število ponovno objavljenih vsebin | Opombe | Provider | Number of cases of user complaints | Number of reinstated accounts | Number of reinstated contents | Comments |
Google | 378 | 0 | Ni na voljo. | Googlovi računi, onemogočeni zaradi spolne zlorabe otrok na spletu, kjer se je uporabnik pritožil: 378. Noben račun ni bil obnovljen. | Google | 378 | 0 | n.a. | Google accounts disabled for online child sexual abuse where the user appealed: 378. None were reinstated. |
LinkedIn | 0 | Ni na voljo. | Ni na voljo. | LinkedIn | 0 | n.a. | n.a. |
Meta | 29 000 | Ni na voljo. | 3 700 | Uporabniki so se pritožili zoper ukrepe za približno 29 000 deljenih medijskih vsebin. Po pritožbenem postopku je bilo približno 3 700 vsebin ponovno objavljenih, ukrepi zoper račune pa razveljavljeni. | Meta | 29 000 | n.a. | 3.700 | Users appealed the actions of around 29 000 pieces of their shared media. Following the appeals process, around 3 700 pieces of content were restored, and account actions reversed. |
Microsoft | 0 | Ni na voljo. | Ni na voljo. | Microsoft | 0 | n.a. | n.a. |
X (prej Twitter) | Pribl. 430 000 | Pribl. 4 000 | Ni na voljo. | Pribl. 430 000 pritožb. Podjetje X je obnovilo pribl. 4 000 od teh računov. | X (former Twitter) | Ca 430 000 | Ca 4 000 | n.a. | Ca 430 000 appeals. X reinstated ca 4 000 of those accounts. |
2.1.6.Število in deleži napak (lažno pozitivnih primerov) pri različnih uporabljenih tehnologijah | 2.1.6.Number and ratios of errors (false positives) of the different technologies used |
Ponudniki morajo v skladu s členom 3(1)(e) začasne uredbe zagotoviti, da so uporabljene tehnologije same po sebi dovolj zanesljive, tako da čim bolj zmanjšujejo stopnjo napake pri odkrivanju vsebin, ki predstavljajo spolno zlorabo otrok na spletu. | In accordance with Article 3(1)(e) of the Interim Regulation, providers are to ensure that the technologies used are sufficiently reliable in that they limit to the maximum extent possible the rate of errors regarding the detection of content representing online child sexual abuse. |
V zvezi s tem so ponudniki poročali, da tehnologij za odkrivanje spolne zlorabe otrok na spletu ne uporabljajo ločeno. Pri odkrivanju spolne zlorabe otrok na spletu namreč uporabljajo večplastni pristop, saj za večjo natančnost združujejo različne tehnologije odkrivanja. Da bi zmanjšali delež napak ali lažno pozitivnih primerov, vsi ponudniki te tehnologije dopolnjujejo s človeških pregledom. Ponudniki niso sporočili števila ali deležev napak (lažno pozitivnih primerov) za vsako od različnih tehnologij, uporabljenih ločeno, ampak so sporočili zbirne podatke za vse uporabljene tehnologije. | In this respect, providers reported that they do not apply each technology to detect online CSA in isolation. Rather, they implement a layered approach to detection of online CSA by combining different detection technologies to increase accuracy. To reduce errors or false positives, all providers complement these with human review. Providers did not provide the number and ratios of errors (false positives) for each of the different technologies used separately, but rather reported aggregate data for all technologies used. |
Večina ponudnikov meri število ali deleže napak kot razveljavljene sklepe o izvrševanju, tj. kot splošno stopnjo obnovitve/razveljavitve po pritožbi (npr. stopnja, po kateri je ponudnik obnovil onemogočene račune ali ponovno objavil vsebino po pritožbi uporabnika). Pristop, ki so ga uporabili ponudniki, ne odraža nujno opredelitve „lažno pozitivnih primerov“ v statističnih podatkih. | Most providers measure the number and ratios of errors as enforcement decision reversals, i.e. as the overall appeal reinstate/reversal rate (e.g. the rate at which the provider reinstated disabled accounts or content upon the user appeal). The approach taken by the providers does not necessarily reflect the definition of ‘false positives’ in statistics. |
Sporočene stopnje obnovitve/razveljavitve po pritožbi so naslednje: | The appeal reinstate/reversal rates reported are as follows: |
Preglednica št. 5: Stopnje obnovitve/razveljavitve po pritožbi | Table n.5: Appeal reinstate/reversal rates |
2021* | 2022 | 2021* | 2022 |
Ponudnik | Obnovljeni računi v % glede na število pritožb | Obnovljeni računi v % glede na preklicane račune | Obnovljeni računi v % glede na število pritožb | Obnovljeni računi v % glede na preklicane račune | Opombe | Provider | % of reinstated accounts vs number of appeals | % of reinstated vs suspended accounts | % of reinstated accounts vs number of appeals | % of reinstated vs suspended accounts | Comments |
Google | 0 % (0 od 8) | 0 % (0 od 33) | 0 % (0 od 378) | 0 % | (glej opombe) | Število preklicanih računov za leto 2022 ni bilo sporočeno. Sporočeno je bilo število vsebin, ki so bile odkrite in prijavljene centru NCMEC, tj. 2 045. Po pritožbi ni bil obnovljen noben račun. | Google | 0 % (0 vs 8) | 0 % (0 vs 33) | 0 % (0 vs 378) | 0 % | (see comments) | The number of suspended accounts was not provided for 2022. Instead, it was provided the number of pieces of content identified and reported to NCMEC, 2045. No account was reinstated following the appeal. |
LinkedIn | 0 % | 0 % | 0 % | 0 % | Ni bilo pritožb. Za obdobje od 13. julija 2021 do 31. decembra 2021 je podjetje LinkedIn tudi poročalo, da je bilo od 75 datotek, pregledanih zaradi suma, da gre za posnetke spolne zlorabe otrok, ki so izhajale iz EU, 31 s človeškim pregledom potrjenih kot posnetki spolne zlorabe otrok. Podjetje za leto 2022 ni predložilo takih podatkov. | LinkedIn | 0 % | 0 % | 0 % | 0 % | No appeals. For the period from 13 July 2021 to 31 December 2021, LinkedIn also reported that of the 75 files reviewed as potential CSAM originating from the EU, 31 were confirmed as CSAM by human review. LinkedIn did not provide such data for 2022. |
Meta | 4,22 % | (207 od 4 900) | 0,06 % | (207 od 340 000) | Glej opombe | Glej opombe | Za natančno določitev obsega, vsebine pritožb in razlogov za obnovitev ni bilo predloženih informacij. | Podatki, predloženi za leto 2022, so se nanašali na vsebine in ne na račune: | –Število preklicanih vsebin („zoper katere so bili uvedeni ukrepi“): 6,6 milijonov; | –število preklicanih vsebin, zoper katere je bila vložena pritožba: 29 000; | –število ponovno objavljenih vsebin: 3 700 | Torej: | –ponovno objavljene vsebine v % glede na število pritožb: 12,8 % (3 700 od 29 000) | –ponovno objavljene vsebine v % glede na število preklicanih vsebin: 0,06 % (3 700 od 6,6 milijona). | Meta | 4.22 % | (207 vs 4.9k) | 0.06 % | (207 vs 340k) | See comments | See comments | No information was provided to determine precisely the scope, the content of the appeals and the reasons to reinstate. | For 2022, the data provided was in terms of pieces of content, not in terms of accounts: | -Number of pieces of content suspended (“actioned”): 6.6 million | -Number of pieces of content appealed: 29k | -Number of pieces of content reinstated: 3.7k | Therefore: | -% of reinstated pieces of content vs number of appeals: 12.8% (3.7k vs 29k) | -% of reinstated pieces of content vs number of pieces of content suspended: 0.06% (3.7k vs 6.6 million) |
Microsoft | 0 % | – | – | – | Ni dovolj podatkov, da bi bilo mogoče izračunati stopnjo obnovitve/razveljavitve po pritožbi. Leta 2022 je bilo 17 popolnih razveljavitev prvotnih odločitev o moderiranju vsebine, podatki o skupnem številu pritožb pa niso bili predloženi. | Microsoft | 0 % | - | - | - | Insufficient data to enable the calculation of the appeal reinstate/reversal rate. For 2022 17 total reversals of initial content moderation decision, no figures on total appeals provided. |
X (prej Twitter) | 1,43 % | (100 od 7 000) | 0,06 % | (100 od 166 000) | 2,17 % | (500 od 23 000) | 0,10 % | (500 od 501 000) | V drugi polovici leta 2021 je bilo z avtomatiziranimi mehanizmi zaradi spolne zlorabe otrok preklicanih približno 166 000 računov uporabnikov. Od teh uporabnikov jih je približno 7 000 vložilo pritožbe, na podlagi katerih je bilo izrečenih približno 100 razveljavitev. | Leta 2022 je bilo z avtomatiziranimi mehanizmi zaradi spolne zlorabe otrok preklicanih 501 000 računov uporabnikov. Od teh uporabnikov jih je približno 23 000 vložilo pritožbe, na podlagi katerih je bilo izrečenih približno 500 razveljavitev. | X (former Twitter) | 1.43% | (100 vs 7k) | 0.06 % | (100 vs 166k) | 2.17% | (500 vs 23k) | 0,10% | (500 vs 501k) | For the second half of 2021 ca 166k users suspended for CSA via automated mechanisms. Of these users, ca 7k submitted appeals which resulted in ca 100 overturns. | In 2022, 501k users suspended for CSA via automated mechanisms. Of these users, ca 23k submitted appeals which resulted in ca 500 overturns. |
* Obdobja poročanja za leto 2021 se razlikujejo glede na ponudnika. | * The reporting periods in 2021 vary for each provider |
2.1.7.Ukrepi, sprejeti za omejitev stopnje napak, in dosežena stopnja napak | 2.1.7.Measures applied to limit the error rate and the error rate achieved |
V skladu s členom 3(1)(e) začasne uredbe morajo biti uporabljene tehnologije same po sebi dovolj zanesljive, posledice napak, če do njih vendarle pride, pa je treba brez odlašanja odpraviti. Poleg tega člen 3(1)(g)(ii) zahteva človeški nadzor in po potrebi človekovo posredovanje. | According to Article 3(1)(e) of the Interim Regulation, the technologies used must be sufficiently reliable and the consequences of any occasional errors must be rectified without delay. In addition, Article 3(1)(g)(ii) requires human oversight and, where necessary, human intervention. |
Vsi ponudniki so poročali, da uporabljajo večplastni pristop za odkrivanje spolne zlorabe otrok na spletu in za boj proti njenemu širjenju. To vključuje uporabo tehnologij za iskanje ujemanj edinstvenega, nerekonvertibilnega digitalnega podpisa (vključno s PhotoDNA) za odkrivanje posnetkov spolne zlorabe otrok v kombinaciji s postopki človeškega pregleda, da se potrdi, ali medijska datoteka (podoba ali videoposnetek) vsebuje posnetke spolne zlorabe otrok, ter tudi človeški nadzor nad obdelavo posnetkov spolne zlorabe otrok. | All providers reported using a layered approach to detect and combat the spread of online CSA. This includes the use of hash-matching technologies (including PhotoDNA) to detect CSAM in combination with human review processes to confirm whether a media file (image and video) contains CSAM, as well as human oversight over the CSAM processing. |
Ponudniki so poročali, da so uporabljali različne zaščitne in druge ukrepe, da bi omejili in zmanjšali stopnjo napak pri odkrivanju in odstranjevanju spolne zlorabe otrok na spletu ter njenem prijavljanju. Ti vključujejo (neizčrpen seznam) 4 : | Providers reported applying different measures and safeguards to limit and reduce the error rate in their detection, reporting and removal of online CSA. These include (non-exhaustive list) 4 : |
I.spremljanje in ocenjevanje kakovosti delovanja orodij za odkrivanje spolne zlorabe otrok, in sicer za povečanje natančnosti (da odkrivajo le spolno zlorabo otrok na spletu) in priklic (da na svojih platformah ne spregledajo spolne zlorabe otrok na spletu) (Google, X); | I.monitoring and quality assessment of the performance of CSA detection tools, both to fine tune precision (that they are detecting only online child sexual abuse) and recall (that they are not missing online child sexual abuse on their platforms) (Google, X); |
II.človeški pregled in nadzor: vzorce medijev, ki jih tehnologije za iskanje ujemanj edinstvenega, nerekonvertibilnega digitalnega podpisa odkrijejo kot posnetke spolne zlorabe otrok, pregledajo pregledovalci/usposobljeni analitiki (Google, LinkedIn, Meta, Microsoft); | II.human review and oversight: samples of media detected as CSAM by hash-matching technologies are audited by human reviewers/trained analysts (Google, LinkedIn, Meta, Microsoft); |
III.označevanje in pregled obsežnih sklopov (Meta); | III.flagging and review of high-volume clusters (Meta); |
IV.uporabo nadaljnjih postopkov ročnega pregleda kot stalnih preverjanj kakovosti edinstvenega, nerekonvertibilnega digitalnega podpisa (LinkedIn, Microsoft); | IV.deployment of further manual review processes as ongoing hash quality checks (LinkedIn, Microsoft); |
V.pregledovalci opravijo specializirana intenzivna usposabljanja pod nadzorom svetovalca v zvezi s tem, kako prepoznati vsebino posnetkov spolne zlorabe otrok, da se zagotovi točnost človeškega pregleda (Google); | V.human reviewers undergoing specialised robust trainings under guidance of counsel on how to recognise CSAM content to ensure accuracy of human review (Google); |
VI.redna ocenjevanja nadzora kakovosti pregledovalcev in izrečenih obsodb (Google, X); | VI.periodic quality control assessments of human reviewers and the verdicts that are applied (Google, X); |
VII.druge postopke za nadzor kakovosti, da se zmanjšajo napake in, če se pojavijo, nemudoma odpravijo, kot sta neodvisno preverjanje edinstvenega, nerekonvertibilnega digitalnega podpisa (Google, LinkedIn) in človeški pregled vsakega primera še nikoli videnega posnetka spolne zlorabe otrok pred prijavo (Google); | VII.other quality control processes to reduce errors and immediate remedy, such as independent hash verification (Google, LinkedIn), human review of each instance of never-before-seen CSAM prior to reporting (Google); |
VIII.priprava in redni pregled politik in strategij izvrševanja s strani strokovnjakov, usposobljenih na področju spolne zlorabe otrok na spletu (Google); | VIII.development and regular review of policies and enforcement strategies by trained subject matter experts on online CSA (Google); |
IX.sodelovanje pri zagotavljanju kakovosti poročil v sistemu CyberTipline centra NCMEC in lažnih pozitivnih primerih, če obstajajo. (Google, LinkedIn, Meta, Microsoft, X) | IX.engagement with NCMEC CyberTipline reports quality, and false positives, if any. (Google, LinkedIn, Meta, Microsoft, X). |
2.1.8.Politika hrambe podatkov in ukrepi, sprejeti za zaščito podatkov na podlagi splošne uredbe o varstvu podatkov | 2.1.8.The retention policy and the data protection safeguards applied pursuant to the GDPR |
Člen 3(1)(h) in (i) začasne uredbe določa, da se ustrezni osebni podatki varno shranijo le za določene posebne namene oziroma da vsebujejo specifikacije v zvezi z obdobjem hrambe. Poleg tega je treba upoštevati veljavne zahteve splošne uredbe o varstvu podatkov. | Article 3(1)(h) and (i) of the Interim Regulation require relevant personal data to be stored in a secure manner only for certain specified purposes and contain specifications regarding the storage period, respectively. In addition, the applicable requirements of the GDPR must be respected. |
Ponudniki so poročali, da imajo vzpostavljene trdne politike hrambe podatkov in zaščitne ukrepe za varstvo osebnih podatkov. Politike hrambe podatkov se razlikujejo glede na vrsto podatkov. Navajajo, da je v vsakem primeru obdobje hrambe časovno omejeno, kot se zdi ustrezno za vrsto podatkov in namen obdelave, ter da se podatki ob koncu obdobja hrambe izbrišejo Ponudniki imajo informacije o praksah hrambe podatkov podrobneje opredeljene v svojih politikah zasebnosti/izjavah o zasebnosti in pogodbah o storitvah/uporabi. | Providers reported having robust retention policies and personal data protection safeguards in place. Data retention policies vary depending on the type of data. They indicate that in each case the retention period is limited in time as deemed appropriate for the type of data and the purpose of processing and the data is deleted at the end of the retention period. Providers have more detailed information on data retention practices defined in their Privacy Policy/Statements and Service/User Agreements. |
Večina ponudnikov (LinkedIn, Meta, Microsoft) uporablja 90-dnevno obdobje hrambe za medije, ki potrjeno vsebujejo posnetke spolne zlorabe otrok, odkrite v okviru področja uporabe Uredbe. V tem obdobju se vsebina, ki je potrjena kot posnetki spolne zlorabe otrok, hrani v ločeni in varni hrambi za posnetke spolne zlorabe otrok, ki jo upravljajo specializirane ekipe (npr. Microsoftova ekipa za kazenski pregon in nacionalno varnost). Ti sistemi hrambe po 90 dneh samodejno izbrišejo shranjeno vsebino s posnetki spolne zlorabe otrok, razen če je bilo obdobje hrambe podaljšano po prejemu zahtevkov za sodni postopek, ki so načeloma povezani z organi kazenskega pregona, ki nadalje ukrepajo na podlagi prijav centra NCMEC. | Most providers (LinkedIn, Meta, Microsoft) apply a retention policy of 90 days for media confirmed to contain CSAM detected within the scope of the Regulation. During this period, the content confirmed as CSAM is stored in separate and secure CSAM storage managed by specialised teams (e.g. Microsoft’s Law Enforcement and National Security team). These storage systems automatically delete the stored CSAM content after 90 days unless the storage period has been extended upon receipt of lawful process requests generally related to law enforcement agencies following up on NCMEC reports. |
Podjetje Google je poročalo, da se posnetki spolne zlorabe otrok, odkriti v okviru področja uporabe Uredbe, hranijo toliko časa, kot je nujno potrebno za zadevne namene iz Uredbe, v vsakem primeru pa ne dlje kot 12 mesecev od datuma, ko so bili posnetki spolne zlorabe otrok odkriti in sporočeni, pri čemer se obdobje lahko podaljša na podlagi veljavnega pravnega zahtevka za ohranitev. | Google reported that CSAM detected within the scope of the Regulation is stored no longer than strictly necessary for the relevant purposes under the Regulation and, in any event, no longer than 12 months from when the CSAM is identified and reported, with a possible extension based on a valid legal preservation request. |
Podjetje X (prej Twitter) je poročalo, da informacije in vsebino profilov hrani za čas trajanja uporabniškega računa, osebne podatke, zbrane od uporabnikov, ko ti uporabljajo njihove storitve, pa največ 18 mesecev. Če uporabnik deaktivira račun, podjetje X običajno hrani podatke še dodatnih 30 dni, nato pa račun izbriše. Podatki uporabnika, povezani s pritožbami in kršitvami politike, vključno z informacijami o računih kršiteljev (npr. identifikatorji, uporabljeni za ustvarjanje računa: naslov e-pošte ali telefonska številka), se hranijo za nedoločen čas, da bi se večkratnim kršiteljem politike preprečilo ustvarjanje novih računov in zagotovilo, da kršitelji politik podjetja X ne morejo preprosto počakati do obdobja izbrisa in nato ponovno kršiti politike 5 . | X (former Twitter) reported keeping profile information and content for the duration of the user account and personal data collected when users use their service for a maximum of 18 months. When an account is deactivated by the user, X generally keeps the data for an additional 30 days, then the account will go for deletion. User data related to complaints and policy violations, including the account information of violators (e.g. identifiers used to create the account: e-mail address or phone number), is retained indefinitely to prevent repeat policy offenders from creating new accounts and ensure that violators of X’s policies cannot simply wait for the deletion period and then violate policies again. 5 |
Zaščitni ukrepi za varstvo osebnih podatkov, ki so jih izvedli ponudniki, vključujejo ukrepe, standardne za panogo (vse ali izbor ukrepov), kot so (neizčrpen seznam) 6 : | Personal data protection safeguards implemented by providers include industry standard measures (all or selection of), such as (non-exhaustive list) 6 : |
I.uporaba tehnik deidentifikacije ali psevdonimizacije in anonimizacije podatkov (npr. prikrivanje, zgoščevanje, diferencirana zasebnost) (Goggle, LinkedIn, Meta, Microsoft); | I.Use of de-identification or pseudonymisation techniques and anonymisation of data (e.g. masking, hashing, differential privacy) (Goggle, LinkedIn, Meta, Microsoft); |
II.zagotavljanje le zgoščenih vrednosti tretjim osebam za namen odkrivanja posnetkov spolne zlorabe otrok (Google, LinkedIn); | II.Provision of only hash values to third parties for the purpose of CSAM detection (Google, LinkedIn); |
III.uporaba za panogo standardnega šifriranja (algoritmi in protokoli) za podatke v tranzitu med infrastrukturo v zasebni lasti in javnimi omrežji (Meta); | III.Use of industry standard encryption (algorithms and protocols) for data in transit between privately owned infrastructure and public networks (Meta); |
IV.izvajanje strategij za upravljanje podatkov/celovitih programov zasebnosti (X (prej Twitter), Google) in strogih notranjih omejitev za dostop do podatkov (Meta) (npr. uporabljajo se za osebje, pogodbenike in posrednike, ki potrebujejo informacije za njihovo obdelavo) ter uporaba kontrolnih seznamov dostopa v orodjih za pregled posnetkov spolne zlorabe otrok in prepovedi edinstvenega, nerekonvertibilnega digitalnega podpisa (Meta) in strogih pogodbenih obveznosti glede zaupnosti, ki se uporabljajo za tiste z dostopom; | IV.Implementation of data governance strategies/comprehensive privacy programmes (X, former Twitter, Google) and of strict internal data access restrictions (Meta) (e.g. applied to staff, contractors, agents who need the information in order to process it), usage of Access Control Lists across CSAM review tools and hash bans (Meta), and strict contractual confidentiality obligations applied to those with access; |
V.pregled strategij za anonimizacijo in upravljanje podatkov, tj. izvedba pregledov zasebnosti za opredelitev morebitnih tveganj za zasebnost, ki izhajajo iz zbiranja, obdelave, hrambe in izmenjave osebnih podatkov, dostop do njih in njihovo zmanjšanje, pregled praks varstva podatkov (Microsoft); | V.Review of anonymisation and data governance strategies, i.e. conducting privacy reviews to identify, access and mitigate potential privacy risks from the collection, processing, storing and sharing of personal data, review of protection practices (Microsoft); |
VI.vzdrževanje načrtov odzivanja na varnostne incidente za spremljanje, odkrivanje in obravnavanje možnih varnostnih ranljivosti in incidentov v infrastrukturi (Google, Meta). | VI.Maintaining security incident response plans for monitoring, detecting, and handling any possible security vulnerabilities and incidents across infrastructure (Google, Meta). |
2.1.9.Imena organizacij, ki delujejo v javnem interesu proti spolni zlorabi otrok, katerih podatki so bili posredovani na podlagi Uredbe | 2.1.9.The names of the organisations acting in the public interest against child sexual abuse with which data has been shared pursuant to this Regulation. |
V obeh obdobjih poročanja (od julija/avgusta 2021 do decembra 2021 in od januarja 2022 do decembra 2022) so vsi ponudniki poročali o izmenjavi podatkov, obdelanih na podlagi Uredbe, s centrom NCMEC. Vsi ponudniki, ki so poročali, so Komisiji, v skladu s členom 7(1) začasne uredbe, tudi sporočili, da so centru NCMEC na podlagi Uredbe prijavili spolno zlorabo otrok na spletu 7 . | In both reporting periods (July/August 2021 to December 2021 and January 2022 to December 2022), all providers reported sharing the data processed under this Regulation with NCMEC. All reporting providers also communicated to the Commission, in compliance with Article 7(1) of the Interim Regulation, that they reported online child sexual abuse under this Regulation to NCMEC. 7 |
2.2Statistični podatki Komisije (člen 8) | 2.2.Member States' statistics (Article 8) |
Države članice morajo v skladu s členom 8(1) začasne uredbe predložiti statistične podatke o: | Member States are obliged to provide statistics pursuant to Article 8(1) of the Interim Regulation on the following: |
(a)skupnem številu prijav odkritih spolnih zlorabah otrok na spletu, ki so jih pristojni nacionalni organi preprečevanja, odkrivanja in preiskovanja kaznivih dejanj prejeli od ponudnikov in organizacij, ki delujejo v javnem interesu proti spolni zlorabi otrok, pri čemer se v primeru, da so take informacije na voljo, razlikuje med absolutnim številom primerov in večkrat prijavljenimi primeri ter vrsto ponudnika, pri storitvi katerega je bila odkrita spolna zloraba otrok na spletu; | (a)the total number of reports of detected online child sexual abuse that have been submitted by providers and organisations acting in the public interest against child sexual abuse to the competent national law enforcement authorities, differentiating, where such information is available, between the absolute number of cases and those cases reported several times and the type of provider on whose service the online child sexual abuse was detected; |
(b)številu otrok, identificiranih z ukrepi na podlagi člena 3, glede na spol; | (b)the number of children identified through actions pursuant to Article 3, differentiated by gender; |
(c)številu obsojenih storilcev. | (c)the number of perpetrators convicted. |
Večina držav članic je sicer zagotovila vsaj delne informacije, vendar veliko držav članic ni vzpostavilo ustreznih sistemov zbiranja in sporočanja podatkov. Zato se statistični podatki, ki so bili predloženi, nanašajo na zelo različna obdobja poročanja, zelo pa se razlikujejo tudi z vidika podrobnosti. Nekatere države članice so predložile letne statistične podatke od datuma začetka veljavnosti Uredbe. Večina jih je poročala za koledarska leta, saj morda nimajo tehničnih sredstev za razlikovanje med zahtevanimi statističnimi podatki po letih od datuma začetka veljavnosti Uredbe. Nekaj držav članic sploh ni predložilo podatkov. | While most Member States provided at least partial information, the relevant data collection and reporting systems had not been set up in many of the Member States. As a result, where statistics were provided, they relate to very diverse reporting periods and differ significantly in terms of granularity. Some Member States submitted yearly statistics as of the date of entry into force of the Regulation. Most of them reported for calendar years as they might not have the technical means to distinguish the requested statistics per year as of the date of entry into force of the Regulation. A few Member States provided no data at all. |
Opozoriti je tudi treba, da se v nekaterih primerih statistični podatki pridobivajo iz tako imenovanih tekočih podatkovnih zbirk ali sistemov dokumentiranja in vodenja zadev, tj. ne iz dejanskih sistemov statističnih podatkov. Podatki se občasno predložijo na podlagi dinamičnih podatkov, kar pomeni, da podatki niso dokončni, tj. se spreminjajo. Spremembe se izvedejo, npr. odvisno od časa pridobivanja (npr. v Sloveniji in na Danskem), ko se zaključi več preiskav in sodnih zadev. | It should also be noted that in some cases the statistical data are obtained from so-called current databases, or journalisation and case management systems, i.e. not actual statistics systems. The numbers are at times provided based on dynamic data, which means the data are not final, i.e. they are subject to change. Changes occur, e.g., depending on the time of extraction (e.g. in Slovenia and Denmark) as more investigations and court cases are completed. |
V več državah članica pristojni organi ustanavljajo nove oddelke za preiskovanje kaznivih dejanj, povezanih s spolno zlorabo otrok na spletu, in vzpostavljajo centralno poročanje za spolno zlorabo otrok na spletu (Latvija in Češka). To bi moralo prispevati k natančnejšim statističnim podatkov v prihodnje. | In several Member States, the competent authorities are creating new departments for investigating crimes related to child sexual abuse online, and creating central reporting for child sexual abuse online (Latvia, Czechia). This should help in having more accurate statistics in the future. |
Nemčija je navedla, da ni mogla predložiti statističnih podatkov v skladu s členom 8(1) začasne uredbe, saj je menila, da nima pravne podlage za prostovoljno odkrivanje 8 . Vendar nemški zvezni kriminalistični urad (BKA) na svojem spletišču poroča, da je leta 2022 od centra NCMEC prejel 89 844 prijav, center NCMEC pa poroča, da je nemškim organom poslal 138 193 prijav 9 . Tri države članice niso predložile nobenih podatkov ali utemeljitve, zakaj niso poročale, v skladu z navedeno določbo (Malta, Portugalska in Romunija). | Germany stated that it could not provide any statistics according to Article 8(1) of the Interim Regulation, as it considered it had no legal basis for voluntary detection. 8 However, the German Federal Criminal Office (BKA) on its website reports receipt of 89 844 reports from NCMEC in 2022, and NCMEC reports sending 138 193 reports to the German authorities. 9 Three Member States did not provide any data or justification for not reporting pursuant to that provision (Malta, Portugal and Romania). |
2.2.1 Skupno število prijav odkritih spolnih zlorab otrok na spletu | 2.2.1. The total number of reports of detected online child sexual abuse |
Večina držav članic je predložila nekaj statističnih podatkov o skupnem številu prijav spolnih zlorab otrok na spletu v skladu s členom 8(1)(a) začasne uredbe. Ker so države članice predložila podatke za različna obdobja poročanja, ni bilo mogoče izračunati skupnega števila prijav odkritih spolnih zlorab otrok na splet, prejetih na ravni EU za katero koli obdobje, kot je čas izvajanja Uredbe. | Most Member States provided some statistics on the total number of reports of online child sexual abuse pursuant to Article 8(1)(a) of the Interim Regulation. As Member States provided data for differing reporting periods, it was not possible to calculate the total number of reports of detected online child sexual abuse received at EU level for any given period such as the time of implementation of the Regulation. |
Države članice so večinoma navedle skupno število prijav, ki so jih nacionalni organi kazenskega pregona prejeli od ponudnikov ali drugih organizacij, ki delujejo v javnem interesu proti spolni zlorabi otrok. Glede na to, da večina ponudnikov s sedežem v ZDA poda prijave centru NCMEC, je večina držav članic poročala, da so večino prijav ali vse prijave prejele od centra NCMEC. Države članice niso navedle števila prijav, na podlagi katerih je mogoče ukrepati, tj. prijav, primernih za preiskavo, so pa nekatere navedle število sproženih postopkov, ki je precej nižje. Razlika med prejetimi prijavami in preiskanimi primeri je bila pripisana več razlogom, npr. da je prijava vsebovala posnetke spolne zlorabe otrok, ni pa vsebovala dovolj informacij za začetek preiskave; združevanje prijav, kadar se več prijav nanaša na določenega osumljenca, ali da gradivo, čeprav prikazuje izkoriščanje, po nacionalnem pravu ni bilo ocenjeno kot kaznivo dejanje. Poleg tega države članice večinoma niso razlikovale med absolutnim številom primerov in večkrat prijavljenimi primeri. V primeru prijav, o katerih je poročal center NCMEC, je center prijave, ki jih je prejel od ponudnikov, že vnaprej razvrstil kot „prijave, na podlagi katerih je mogoče ukrepati“, ali kot „informativne prijave“. Center NCMEC prijavo, na podlagi katere je mogoče ukrepati, opredeljuje kot prijavo, ki vsebuje dovolj informacij za začetek preiskave. To običajno zajema podrobnosti o uporabniku, posnetke in možno lokacijo. Prijava je razvrščena kot „informativna“, če ne vsebuje dovolj informacij ali če posnetki veljajo za viralne in so bili večkrat prijavljeni. Center NCMEC je leta 2022 49 % prijav označil kot „prijave, na podlagi katerih je mogoče ukrepati“, 51 % prijav pa kot „informativne“. | Member States mostly provided the total number of reports received from providers or other organisations acting in the public interest against child sexual abuse to the national law enforcement authorities. Given that most US-based providers report to NCMEC, most Member States reported receiving most or all of their reports from NCMEC. Member States did not indicate the number of actionable reports, i.e. reports suitable for investigation, but some indicated the number of cases launched, which is significantly lower. The difference between reports received and investigated cases were attributed to several reasons, e.g. that the report contained CSAM, but it did not contain sufficient information to open an investigation; the merging of reports when more than one report applies to a certain suspect; or that the material, while showing exploitative situations, was not assessed as criminal under national law. In addition, Member State mostly did not differentiate the absolute number of cases and those reported several times. Where reports were provided by NCMEC, NCMEC already pre-categorised reports it had received from providers as “actionable” or “informational”. NCMEC defines an actionable report as one containing sufficient information to launch an investigation. This typically includes user details, imagery, and a possible location. The report is categorised as “informational” when it contains insufficient information or where the imagery is considered viral and has been reported many times. NCMEC designated 49% of the reports as “actionable” in 2022, while 51% were designated as “informational”. |
Le nekaj držav članic je navedlo vrsto ponudnikov, pri storitvah katerih je bila odkrita spolna zloraba otrok na spletu (npr. Belgija, Češka, Estonija, Francija in Poljska), le ena država članica pa je predložila podrobno razčlenitev (Belgija). | Only very few Member States indicated the type of providers on whose services the online child sexual abuse was detected (e.g. Belgium, Czechia, Estonia, France, and Poland) and only one Member State provided a detailed breakdown (Belgium). |
Slovenija je navedla, da ni mogla predložiti podatkov le o kaznivih dejanjih, preiskanih zaradi prijav ponudnikov in organizacij, ampak je lahko predložila le podatke za vse preiskave spolne zlorabe otrok na spletu ne glede na vir informacij, na podlagi katerih se je začela preiskava. | Slovenia indicated that it could not provide figures only on offences investigated because of reports submitted by providers and organisations, but rather could only provide figures for all investigations into online child sexual abuse regardless of the source of the information that led to the launch of the investigation. |
Preglednica št. 6: Skupno število prijav odkritih spolnih zlorab otrok na spletu, kot so jih sporočile države članice | Table n.6: The total number of reports of detected online child sexual abuse as reported by Member States |
Država | Obdobje poročanja | Skupno število prijav spolnih zlorab otrok na spletu | Vir prijav | Opombe | Country | Reporting Period | Total number of reports of online child sexual abuse | Source of reports | Comments |
Avstrija | 2021–2022 | 16 311 | Center NCMEC | Austria | 2021 to 2022 | 16 311 | NCMEC |
Belgija | 1. avgust 2021–31. julij 2022 | 26 226 | Prijave, ki so jih podali ponudniki (družbeni mediji) in prijavna točka Child Focus | Belgium | 1 August 2021 to 31 July 2022 | 26 226 | Reports originating from providers (social media) and Childfocus hotline |
Bolgarija | 2021–2022 | 9 120 | Ponudniki ter prijavna točka INHOPE prek sistema „Safenet“ in drugi | Od tega je bilo 9 112 opozoril o spletnih straneh, ki vsebujejo posnetke spolne zlorabe otrok, ki jih gostijo bolgarski ponudniki. | Bulgaria | 2021 to 2022 | 9 120 | Providers and INHOPE hotline through “Safenet” and other | Out of these, 9 112 alerts were about webpages containing CSAM, hosted by Bulgarian providers. |
Hrvaška | 1. januar 2021–31. oktober 2022 | 9 044 | Center NCMEC | Croatia | 1 January 2021 to 31 October 2022 | 9 044 | NCMEC |
Ciper | 1. julij 2021–31. december 2022 | 3 570 | Center NCMEC | Cyprus | 1 July 2021 to 31 December 2022 | 3 570 | NCMEC |
Češka | 1. januar 2022 do 31. julij 2022 | 13 279 | Center NCMEC | Czechia | 1 January 2022 to 31 July 2022 | 13 279 | NCMEC |
Danska | 2. avgust 2021–20. januar 2023 | 10 744 | Center NCMEC | Denmark | 2 August 2021 to 20 January 2023 | 10 744 | NCMEC |
Estonija | – | – | Center NCMEC, telefonska številka za pomoč otrokom v stiski 116 111 | Estonija je sporočila, da statistični podatki policije in mejne straže, vključno s centrom NCMEC, niso javni. Za leto 2021 so prijavili 360 kaznivih dejanj zoper spolno nedotakljivost otrok brez stika. Poleg tega je bilo 86 % vseh kaznivih dejanj zoper spolno nedotakljivost brez stika storjenih na internetu ali s pomočjo orodij informacijske tehnologije. | Estonia | - | - | NCMEC, Child Helpline 116 111 | Estonia reported that the statistics of Police and Border Guard, including NCMEC, are not public. For 2021 they reported 360 non-contact sexual crimes against a child. Moreover, 86 % of all non-contact sexual crimes were committed in the Internet environment or using information technology tools. |
Finska | 2022 | 25 000 | Center NCMEC in organizacija Save the Children | Finland | 2022 | 25 000 | NCMEC and Save the Children |
Francija | 1. avgust 2021–1. avgust 2022 | 120 000 | Center NCMEC | France | 1 August 2021 to 1 August 2022 | 120 000 | NCMEC |
Nemčija | – | – | – | Podatki niso na voljo/niso bili sporočeni. | Germany | - | - | - | Data not available/not reported. |
Grčija | 2021–2022 | 142 | Center NCMEC, grška prijavna točka za prijavo nezakonitih vsebin na internetu – Safeline, nacionalna komisija za telekomunikacije in pošto, nacionalna telefonska številka za klic v sili 1056 – The smile of the Child, grški varuh človekovih pravic | Greece | 2021 to 2022 | 142 | NCMEC, Greek Hotline for illegal Internet content – Safeline, National Telecommunications and Postal Commission, National Line SOS 1056 – The smile of the child, Greek Ombudsman |
Madžarska | 2022 | 0 | Nobena od prijav, ki so jih poslali ponudniki, ni bila podana na podlagi začasne uredbe. | Hungary | 2022 | 0 | None of the reports sent by providers were sent under the Interim Regulation |
Irska | 2021–2022 | 15 355 | Center NCMEC | Ireland | 2021 to 2022 | 15 355 | NCMEC |
Italija | 2022 | 4 607 | Ni navedeno. | Italy | 2022 | 4 607 | Not specified |
Latvija | 1. avgust 2022–6. marec 2023 | Približno 115 do 220 prijav mesečno. | Prijave so poslali nelatvijski ponudniki in organizacije, ki delujejo v javnem interesu proti spolni zlorabi otrok (zlasti center NCMEC), ter latvijski ponudniki in organizacije (zlasti latvijski center za varnejši internet). | Latvia | 1 August 2022 to 6 March 2023 | Approximately 115 to 220 reports monthly | From non-Latvian providers and organisations acting in the public interest against child sexual abuse (mostly NCMEC) and Latvian providers and organisations (mostly Latvian Safer Internet Center) |
Litva | 1. januar 2021–30. junij 2022 | 4 142 | Ni navedeno | Lithuania | 1 January 2021 to 30 June 2022 | 4 142 | Not specified |
Luksemburg | 2021–2022 | 2 491 | Ni navedeno. | Luxembourg | 2021 to 2022 | 2 491 | Not specified |
Malta | – | – | – | Podatki niso bili predloženi/sporočeni. | Malta | - | - | - | Data not submitted/reported. |
Nizozemska | 2021 | 36 537 | Ponudniki in organizacije, ki delujejo v javnem interesu proti spolni zlorabi otrok. | Netherlands | 2021 | 36 537 | Providers and organisations acting in the public interest against child sexual abuse |
Poljska | 3. avgust 2021–3. avgust 2023 | 227 | Ponudniki in organizacije, ki delujejo v javnem interesu proti spolni zlorabi otrok. | Za obdobje od 3. avgusta 2022 do 3. avgusta 2023 je Poljska prejela eno prijavo o pridobivanju otrok za spolne namene in 105 prijav o posnetkih spolne zlorabe otrok. | Poland | 3 August 2021 to 3 August 2023 | 227 | Providers and organisations acting in the public interest against child sexual abuse | For the period from 3 August 2022 to 3 August 2023 Poland noted 1 report of child grooming and 105 reports of CSAM. |
Portugalska | – | – | – | Podatki niso bili predloženi/sporočeni. | Portugal | - | - | - | Data not submitted/reported. |
Romunija | – | – | – | Podatki niso bili predloženi/sporočeni. | Romania | - | - | - | Data not submitted/reported. |
Slovaška | 1. avgust 2021–31. julij 2022 | 7 206 | Ponudniki in organizacije, ki delujejo v javnem interesu proti spolni zlorabi otrok. | Slovakia | 1 August 2021 to 31 July 2022 | 7 206 | Providers and organisations acting in the public interest against child sexual abuse |
Slovenija | 1. januar 2021–14. julij 2023 | 452 | To število se nanaša na kazniva dejanja, povezana z dejavnostmi na internetu. Trenutno Slovenija na podlagi obstoječih statističnih podatkov ne more ločevati med statističnimi podatki o kaznivih dejanjih, preiskanih na podlagi prijav, ki so jih podali ponudniki in organizacije, in statističnimi podatki o drugih prijavah. | Slovenia | 1 January 2021 to 14 July 2023 | 452 | This number indicates criminal offences related to activities on internet. At present, the existing statistical data do not allow Slovenia to separate statistical data on offences investigated on the basis of reports submitted by providers and organisations from statistical data on other reports. |
Španija | 2022 | 31 474 | Organizacije, ki delujejo v javnem interesu proti spolni zlorabi otrok. | Spain | 2022 | 31 474 | Organisations acting in the public interest against child sexual abuse |
Švedska | Avgust 2021–31. december 2022 | 32 830 | Večinoma center NCMEC | Sweden | August 2021 to 31 December 2022 | 32 830 | Mostly NCMEC |
Glede na to, da je center NCMEC glavni vir prijav, je koristno primerjati podatke o prijavah, ki so jih prejele države članice, s podatki, ki jih je predložil center NCMEC o prijavah, poslanih državam članicam. Center NCMEC je leta 2021 prejel skupaj 29 397 681 prijav iz panoge po vsem svetu, od tega jih je 99,7 % (ali 29 309 106) vsebovalo eno ali več podob ali videoposnetkov spolne zlorabe otrok, 0,15 % (ali 44 155) se jih je nanašalo na pridobivanje otrok za spolne namene, 0,05 % (ali 16 032) pa na trgovanje z otroki za namene spolnega izkoriščanja. Leta 2022 je center NCMEC prejel skupaj 32 059 029 prijav, od tega se jih je 99,5 % (ali 31 901 234) nanašalo na podobe ali videoposnetke spolne zlorabe otrok, 0,25 % (ali 80 524) na pridobivanje otrok za spolne namene in 0,06 % (ali 18 336) na trgovanje z otroki za namene spolnega izkoriščanja. Za EU je razčlenitev naslednja: | Given that NCMEC is the main source of reports, it is informative to compare the figures on reports received by Member States to those provided by NCMEC on reports sent to Member States. NCMEC received a total of 29 397 681 reports from industry globally in 2021, of which 99.7% (or 29 309 106) contained one or more child sexual abuse images or videos, 0.15% (or 44 155) related to grooming and 0.05% (or 16 032) to child sex trafficking. In 2022, NCMEC received a total of 32 059 029 reports, of which 99.5% (or 31 901 234) related to child sexual abuse images or videos, 0.25% (80 524) to grooming and 0.06% (or 18 336) to child sex trafficking. For the EU, the breakdown is as follows: |
Preglednica št. 7: Prijave centra NCMEC o domnevni spolni zlorabi otrok na spletu, poslane državam članicam EU v letih 2021 in 2022 | Table n.7: NCMEC reports of suspected online child sexual abuse provided to EU Member States in 2021 and 2022 |
Država | Skupno število prijav leta 2021 10 | % vseh prijav v EU leta 2021 | Skupno število prijav leta 2022 11 | v % vseh prijav v EU leta 2022 | % prebivalstva EU | Country | Total reports 2021 10 | % of EU total 2021 | Total reports 2022 11 | % of EU total 2022 | % of EU population |
Avstrija | 7 580 | 1,36 % | 18 501 | 1,23 % | 2,00 % | Austria | 7 580 | 1.36 % | 18 501 | 1.23 % | 2.00 % |
Belgija | 15 762 | 2,84 % | 50 255 | 3,34 % | 2,60 % | Belgium | 15 762 | 2.84 % | 50 255 | 3.34 % | 2.60 % |
Bolgarija | 13 584 | 2,44 % | 31 937 | 2,12 % | 1,53 % | Bulgaria | 13 584 | 2.44 % | 31 937 | 2.12 % | 1.53 % |
Hrvaška | 4 744 | 0,85 % | 11 693 | 0,78 % | 0,86 % | Croatia | 4 744 | 0.85 % | 11 693 | 0.78 % | 0.86 % |
Ciper | 2 657 | 0,48 % | 7 361 | 0,49 % | 0,20 % | Cyprus | 2 657 | 0.48 % | 7 361 | 0.49 % | 0.20 % |
Češka | 15 004 | 2,70 % | 61 994 | 4,12 % | 2,36 % | Czechia | 15 004 | 2.70 % | 61 994 | 4.12 % | 2.36 % |
Danska | 5 891 | 1,06 % | 30 215 | 2,01 % | 1,31 % | Denmark | 5 891 | 1.06 % | 30 215 | 2.01 % | 1.31 % |
Estonija | 2 729 | 0,49 % | 6 408 | 0,43 % | 0,30 % | Estonia | 2 729 | 0.49 % | 6 408 | 0.43 % | 0.30 % |
Finska | 6 079 | 1,09 % | 10 904 | 0,73 % | 1,24 % | Finland | 6 079 | 1.09 % | 10 904 | 0.73 % | 1.24 % |
Francija | 98 233 | 17,67 % | 227 465 | 15,13 % | 15,16 % | France | 98 233 | 17.67 % | 227 465 | 15.13 % | 15.16 % |
Nemčija | 79 701 | 14,34 % | 138 193 | 9,19 % | 18,59 % | Germany | 79 701 | 14.34 % | 138 193 | 9.19 % | 18.59 % |
Grčija | 14 616 | 2,63 % | 43 345 | 2,88 % | 2,37 % | Greece | 14 616 | 2.63 % | 43 345 | 2.88 % | 2.37 % |
Madžarska | 31 710 | 5,70 % | 109 434 | 7,28 % | 2,16 % | Hungary | 31 710 | 5.70 % | 109 434 | 7.28 % | 2.16 % |
Irska | 7 327 | 1,32 % | 19 770 | 1,31 % | 1,13 % | Ireland | 7 327 | 1.32 % | 19 770 | 1.31 % | 1.13 % |
Italija | 37 480 | 6,74 % | 96 512 | 6,42 % | 13,32 % | Italy | 37 480 | 6.74 % | 96 512 | 6.42 % | 13.32 % |
Latvija | 1 537 | 0,28 % | 3 688 | 0,25 % | 0,42 % | Latvia | 1 537 | 0.28 % | 3 688 | 0.25 % | 0.42 % |
Litva | 3 509 | 0,63 % | 16 603 | 1,10 % | 0,63 % | Lithuania | 3 509 | 0.63 % | 16 603 | 1.10 % | 0.63 % |
Luksemburg | 2 005 | 0,36 % | 2 004 | 0,13 % | 0,14 % | Luxembourg | 2 005 | 0.36 % | 2 004 | 0.13 % | 0.14 % |
Malta | 750 | 0,13 % | 4 713 | 0,31 % | 0,12 % | Malta | 750 | 0.13 % | 4 713 | 0.31 % | 0.12 % |
Nizozemska | 36 790 | 6,62 % | 57 012 | 3,79 % | 3,96 % | Netherlands | 36 790 | 6.62 % | 57 012 | 3.79 % | 3.96 % |
Poljska | 37 758 | 6,79 % | 235 310 | 15,65 % | 8,41 % | Poland | 37 758 | 6.79 % | 235 310 | 15.65 % | 8.41 % |
Portugalska | 34 415 | 6,19 % | 42 674 | 2,84 % | 2,31 % | Portugal | 34 415 | 6.19 % | 42 674 | 2.84 % | 2.31 % |
Romunija | 32 765 | 5,89 % | 96 287 | 6,40 % | 4,25 % | Romania | 32 765 | 5.89 % | 96 287 | 6.40 % | 4.25 % |
Slovaška | 7 275 | 1,31 % | 39 748 | 2,64 % | 1,21 % | Slovakia | 7 275 | 1.31 % | 39 748 | 2.64 % | 1.21 % |
Slovenija | 3 162 | 0,57 % | 14 795 | 0,98 % | 0,47 % | Slovenia | 3 162 | 0.57 % | 14 795 | 0.98 % | 0.47 % |
Španija | 33 136 | 5,96 % | 77 727 | 5,17 % | 10,60 % | Spain | 33 136 | 5.96 % | 77 727 | 5.17 % | 10.60 % |
Švedska | 19 635 | 3,53 % | 48 883 | 3,25 % | 2,33 % | Sweden | 19 635 | 3.53 % | 48 883 | 3.25 % | 2.33 % |
Skupaj | 555 834 | 1 503 431 | Total | 555 834 | 1 503 431 |
Velika razlika med številom prijav leta 2021 in leta 2022, ki kaže na velik porast prijav leta 2022, je večinoma posledica zmanjšanja prostovoljnega odkrivanja med januarjem in avgustom 2021, ko se začasna uredba še ni uporabljala. | The significant disparities between the number of reports in 2021 and 2022, showing a steep increase of reports in 2022, is due in large part to the decrease in voluntary detection between January and August 2021, when the Interim Regulation did not yet apply. |
Center NCMEC v svojih statističnih podatkih po državah članicah EU ne razlikuje glede na vir prijave, zlasti ali je izhajala iz medosebne komunikacijske storitve, neodvisne od številke. Vendar pa center NCMEC zagotavlja statistične podatke o skupnem številu prijav v zvezi z EU, ki izhajajo iz medosebnih komunikacijskih storitev, neodvisnih od številke. Leta 2021 je 283 265 prijav, ki so se nanašale na države članice, izhajalo iz storitve klepeta, sporočanja ali elektronske pošte, tj. 51 % vseh prijav v zvezi z EU. Dodatnih 164 645 prijav (30 % vseh) je izhajalo iz družbenih medijev ali spletnih igralnih platform, ki imajo lahko vgrajene tudi storitve sporočanja ali klepeta. Leta 2021 se je 3 565 prijav v zvezi z EU nanašalo na pridobivanje otrok za spolne namene. Leta 2022 je 1 015 231 prijav, ki so se nanašale na države članice, izhajalo iz storitve klepeta, sporočanja ali elektronske pošte, tj. 68 % vseh prijav v zvezi z EU. Dodatnih 325 847 prijav (22 % vseh) je izhajalo iz družbenih medijev ali spletnih igralnih platform, ki imajo lahko vgrajene tudi storitve sporočanja ali klepeta. Leta 2022 se je 7 561 prijav v zvezi z EU nanašalo na pridobivanje otrok za spolne namene. Ponovno, razlike v številu prijav iz medosebnih komunikacijskih storitev, neodvisnih od številke, v letih 2021 in 2022 so posledica zmanjšanja prostovoljnega odkrivanja med januarjem in avgustom 2021, ko se začasna uredba še ni uporabljala. | NCMEC does not differentiate in its statistics per EU Member State according to the source of the report, in particular whether it stemmed from a number-independent interpersonal communications service. However, NCMEC does provide statistics about the overall number of reports concerning the EU stemming from number-independent interpersonal communications services. In 2021, 283 265 reports concerning Member States stemmed from a chat, messaging, or email service, that is, 51 % of total reports concerning the EU. An additional 164 645 (30% of the total) of reports stemmed from social media or online gaming platforms, which may also have integrated messaging or chat services. In 2021, 3 565 reports concerning the EU were about grooming. In 2022, 1 015 231 reports concerning to Member States stemmed from a chat, messaging, or email service, that is, 68 % of total reports concerning the EU. An additional 325 847 (22% of the total) of reports stemmed from social media or online gaming platforms, which may also have integrated messaging or chat services. In 2022, 7 561 reports concerning the EU were about grooming. Again, the disparities in the number of reports from number-independent interpersonal communication services in 2021 and 2022 is due to the decrease in voluntary detection between January and August 2021, when the Interim Regulation did not yet apply. |
V številnih primerih se delež prijav po državah članicah v grobem ujema z deležem prebivalstva držav članic glede na prebivalstvo EU kot celoto, kar bi lahko nakazovalo na primerljivo pojavnost spolne zlorabe otrok na spletu v državah članicah. Opazna odstopanja so vidna v primeru Španije in Italije, katerih deleža se zdita nizka v primerjavi z deležem prebivalstva EU v obeh navedenih letih, za deleže prijav za druge države članice pa se zdi, da močno nihajo (npr. za Nemčijo, Poljsko, Nizozemsko in Slovaško). Te spremembe se kot take ne odražajo v poročilih o številu primerov, zato je ponovno težko sklepati o korelaciji med prijavami in preiskavami. | The proportion of reports per Member State roughly matches the proportion of the population of the Member State as compared to the EU population as a whole in many cases, which could point to a comparable incidence of child sexual abuse online across Member States. Notable deviations are visible in relation to Spain and Italy, whose percentages appear low compared to the percentage of the EU population across both years, while the proportion of reports for other Member States appear to fluctuate significantly (e.g. Germany, Poland, Netherlands, Slovakia). These changes are not reflected as such in reports on numbers of cases and it is therefore again difficult to draw conclusions on correlation between reports and investigations. |
Zaradi različnih obdobij poročanja neposredna primerjava ni mogoča, kljub temu pa obstajajo pomembne razlike med statističnimi podatki, ki jih je predložil center NCMEC, in podatki, ki so jih sporočile države članice. Poleg tega podatkov centra NCMEC za države članice ni mogoče v celoti primerjati s podatki, ki jih je zagotovila panoga in so navedeni v prejšnjem oddelku. Ker so nekatere razlike morda posledica prijav spolne zlorabe otrok na spletu, ki izhajajo iz virov, ki niso medosebna komunikacija, bi bila potrebna dodatna analiza, saj je možno tudi, da ponudniki, razen tistih, ki so Komisiji doslej pošiljali poročila, izvajajo ukrepe prostovoljnega odkrivanja v zvezi z EU, glede na seznam ponudnikov, ki podajajo prijave centru NCMEC 12 . Vendar pa dejstvo, da se za večino držav članic zdi, da obstajajo velike razlike med številom prijav, za katere center NCMEC navaja, da jih je poslal državam članicam, in številom prijav, za katere države članice navajajo, da so jih prejele, kaže, da zbiranje in sporočanje podatkov, ki ju izvajajo države članice, nista popolna. | In view of the differing reporting periods, no direct match is possible but nonetheless there are significant disparities between the statistics provided by NCMEC and the figures reported by Member States. In addition, the NCMEC figures for the Member States also cannot fully be matched with those provided by industry reported in the previous section. While some of the differences may be due to reports of child sexual abuse online that come from sources other than interpersonal communications, this would require further analysis, as it is also possible that voluntary detection measures concerning the EU by providers other than those that have submitted reports to the Commission to date is taking place, given the list of providers that are reporting to NCMEC 12 . Nonetheless, the fact that for most Member States there appears to be a significant disparity between the number of reports NCMEC lists as having sent to the Member State, and the number of reports the Member State lists as received, suggests that the Member States’ data collection and reporting is not complete. |
Pri vsaki zgoraj navedeni prijavi centra NCMEC so bili povezani podobe in videoposnetki spolne zlorabe otrok umaknjeni iz obtoka. To je pomembno zlasti za sedanje in nekdanje žrtve spolne zlorabe otrok. Študije so pokazale, da se žrtve zaradi nadaljnjega kroženja podob in videoposnetkov, ki prikazujejo njihovo zlorabo, težje spoprijemajo s psihološkimi posledicam zlorabe, nadaljnje kroženje pa pomeni sekundarno obliko viktimizacije. | For each NCMEC report identified above, the associated images and videos of child sexual abuse were taken down and removed from circulation. This is important in particular for current victims and survivors of child sexual abuse. Studies have shown that the continued circulation of images and videos depicting their abuse limits victims’ ability to overcome the psychological effects of the abuse and creates a secondary form of victimisation. |
2.2.2 Število identificiranih otrok | 2.2.2. The number of children identified |
Večina držav članic je predložila popolne ali delne statistične podatke o številu identificiranih otrok glede na spol, v skladu s členom 8(1)(b) začasne uredbe. Vendar več držav članic ni predložilo nobenih podatkov ali utemeljitve, zakaj niso poročale, v skladu z navedeno določbo. | Most Member States provided complete or partial statistics on the number of children identified, differentiated by gender, pursuant to Article 8(1)(b) of the Interim Regulation. However, several Member States did not provide any data or justification for not reporting pursuant to this provision. |
Nekaj držav članic ni sporočilo nobenih statističnih podatkov ali so sporočile le delne podatke za obdobje poročanja, vendar so navedle razloge za to. Med navedenimi razlogi so bili naslednji: | Several Member States did not report any or reported only partial statistics for the reporting period but provided reasons for this. The reasons provided include: |
–otrok, ki so bili žrtve spolne zlorabe na spletu, ni mogoče prešteti (Francija); | -child victims of online CSA cannot be counted (France); |
–podatki niso na voljo, ker se niso zbirali v okviru nacionalnega zbiranja statističnih podatkov / nacionalni organi niso zabeležili teh statističnih podatkov (Danska in Litva); | -data not available as not collected as part of the national statistical data collection / the national authorities did not register these statistics (Denmark, Lithuania); |
–v nacionalni zbirki statističnih podatkov podatki niso razčlenjeni po spolu (Belgija, Ciper, Češka, Grčija, Irska, Italija, Litva in Nizozemska); | -data not disaggregated by gender in the national statistical data collection (Belgium, Cyprus, Czechia, Greece, Ireland, Italy, Lithuania, Netherlands); |
–v obstoječih informacijskih sistemih informacije niso na voljo z zahtevano ravnjo podrobnosti (Finska); | -information is not available with the requested level of detail in the existing information systems (Finland); |
–informacije se ne zbirajo (Nemčija). | -the information is not collected (Germany). |
Nekatere od držav članic, ki so navedle, da ne morejo predložiti statističnih podatkov, so potrdile, da so bili njihovi nacionalni organi zaprošeni, naj spremenijo postopek registracije za prostovoljne prijave, preiskave in zbiranje statističnih podatkov (Danska) in/ali uporabljajo nove informacijske sisteme, ki bi morali omogočati prijave na zahtevani ravni podrobnosti (Finska). | A few of those Member States that indicated that they were not able to provide statistics confirmed that their national authorities were asked to alter their registration procedure for voluntary reports and investigations and statistics collection (Denmark) and/or are deploying new information systems that should allow reporting at the required level of detail (Finland). |
V eni državi članici pri spodnjih podatkih ni vzpostavljeno razlikovanje med otroki, ki so žrtve spolne zlorabe na spletu, in otroki, ki so žrtve spolne zlorabe zunaj spleta (Madžarska). V nekaterih primerih statistični podatki vključujejo tudi otroke, za katere je bilo ugotovljeno, da so sami ustvarili gradivo in ga prenesli na splet (lastno ustvarjeno gradivo, večinoma videoposnetki) (Češka in Estonija). | In one Member State, the data below do not differentiate between child victims of CSA online and offline (Hungary). In some instances, the statistics also include children that were identified as having produced and uploaded this material themselves (self-generated material, mostly video) (Czechia, Estonia). |
Ker so države članice večinoma poročale za različna obdobja poročanja, ni bilo mogoče izračunati skupnega števila otrok, identificiranih kot žrtve spolne zlorabe na spletu v EU, po letih in/ali istem obdobju poročanja. | As Member States mostly reported for differing reporting periods, it was not possible to calculate the total number of children identified as victims of online child sexual abuse in the EU, per year and/or the same reporting period. |
Preglednica št. 8: Število identificiranih otrok, glede na spol | Table n.8: Number of children identified, differentiated by gender |
Država | Obdobje poročanja | Ženske | Moški | Skupaj | Opombe | Country | Reporting period | Female | Male | Total | Comments |
Avstrija | 2021–2022 | 11 | 6 | 17 | Austria | 2021 to 2022 | 11 | 6 | 17 |
Belgija | 2021–2022 | – | – | 63 | Podatki glede na spol niso na voljo. | Belgium | 2021 to 2022 | - | - | 63 | Data differentiated by gender not available. |
Bolgarija | 2022 | 50 | 12 | 62 | Bulgaria | 2022 | 50 | 12 | 62 |
Hrvaška | 1. januar 2021–31. oktober 2022 | 20 | 0 | 20 | Croatia | 1 January 2021 to 31 October 2022 | 20 | 0 | 20 |
Ciper | 2022 | – | – | 102 | Podatki glede na spol niso na voljo. | Cyprus | 2022 | - | - | 102 | Data differentiated by gender not available. |
Češka | 2022 | – | – | 30 | Podatki glede na spol niso na voljo. | Czechia | 2022 | - | - | 30 | Data differentiated by gender not available. |
Danska | – | – | – | – | Podatki niso na voljo. | Denmark | - | - | - | - | Data not available. |
Estonija | 2021 | 6 | 12 | 18 | Estonia | 2021 | 6 | 12 | 18 |
Finska | – | – | – | – | Podatki niso na voljo. | Finland | - | - | - | - | Data not available. |
Francija | – | – | – | – | Podatki niso na voljo. | France | - | - | - | - | Data not available. |
Nemčija | – | – | – | – | Podatki niso na voljo. | Germany | - | - | - | - | Data not available. |
Grčija | 2021–2022 | – | – | 4 | Podatki glede na spol niso na voljo. | Greece | 2021 to 2022 | - | - | 4 | Data differentiated by gender not available. |
Madžarska | 2021–2022 | 379 | 47 | 426 | Nemogoče je razlikovati med žrtvami spolne zlorabe otrok na spletu in zunaj njega. Vključeni so samo otroci, mlajši od 16 let. | Hungary | 2021 to 2022 | 379 | 47 | 426 | Impossible to differentiate between victims of CSA online and offline. Only children below the age of 16 included. |
Irska | 2021–2022 | – | – | 101 | Podatkov za leto 2021 (50 žrtev) ni mogoče razlikovati glede na spol. Podatki glede na spol za leto 2022: identificiranih je bilo 25 otrok ženskega spola in 26 otrok moškega spola. | Ireland | 2021 to 2022 | - | - | 101 | Data for 2021 (50 victims) cannot be differentiated by gender. Differentiated data by gender for 2022 are: 25 female and 26 male children identified. |
Italija | 2022 | – | – | 385 | Podatki glede na spol niso na voljo. | Italy | 2022 | - | - | 385 | Data differentiated by gender not available. |
Latvija | 1. avgust 2022–6. marec 2023 | 1 | – | 1 | Latvia | 1 August 2022 to 6 March 2023 | 1 | - | 1 |
Litva | – | – | – | – | Podatki niso na voljo. | Lithuania | - | - | - | - | Data not available. |
Luksemburg | 2021–2022 | 0 | 0 | 0 | Luxembourg | 2021 to 2022 | 0 | 0 | 0 |
Malta | – | – | – | – | Podatki niso bili predloženi/sporočeni. | Malta | - | - | - | - | Data not submitted/reported. |
Nizozemska | 2021 | – | – | 222 | Podatki glede na spol niso na voljo. | Netherlands | 2021 | - | - | 222 | Data differentiated by gender not available. |
Poljska | 2022 | 2 368 | 487 | 3 014 | Za leto 2022 poljski nacionalni policijski informacijski sistem zagotavlja podatke o 3 014 žrtvah kaznivih dejanj, povezanih s spolno zlorabo otrok (2 368 žrtev ženskega spola, 487 žrtev moškega spola, 159 žrtev brez navedbe spola). | Poland | 2022 | 2368 | 487 | 3014 | In 2022, data from the National Police Information System in Poland provide the data on 3 014 victims of CSA related offences (2 368 female, 487 male, 159 where the gender is not stated). |
Portugalska | – | – | – | – | Podatki niso bili predloženi/sporočeni. | Portugal | - | - | - | - | Data not submitted/reported. |
Romunija | – | – | – | – | Podatki niso bili predloženi/sporočeni. | Romania | - | - | - | - | Data not submitted/reported. |
Slovaška | Avgust 2021–julij 2022 | 13 | 8 | 21 | Slovakia | August 2021 to July 2022 | 13 | 8 | 21 |
Slovenija | 1. januar 2021–14. julij 2023 | 220 | 85 | 305 | Slovenia | 1 January 2021 to 14 July 2023 | 220 | 85 | 305 |
Španija | 2022 | 80 | 39 | 119 | Spain | 2022 | 80 | 39 | 119 |
Švedska | 2022 | 8 | 4 | 12 | Sweden | 2022 | 8 | 4 | 12 |
SKUPAJ ZA VSE DRŽAVE ČLANICE | 1. januar 2021 do 6. marec 2023 | 3 156 | 700 | 4 922 | TOTAL FOR ALL MEMBER STATES | 1 January 2021 to 6 March 2023 | 3 156 | 700 | 4 922 |
Za zgornje podatke veljajo številni dodatni pridržki. Obstoječi statistični podatki državam članicam ne omogočajo vedno, da bi razlikovale med podatki o žrtvah, ki so bile identificirane na podlagi prijave ponudnika, in podatki o žrtvah, ki so morda same prijavile kaznivo dejanje ali pa je kaznivo dejanje prijavil nekdo drug, ki je žrtev poznal ali je odkril kaznivo dejanje (kot to navaja Slovenija). Švedska je poročala, da so otroci, ki so bili identificirani prek dnevnikov klepeta, prav tako vključeni v poročanje, čeprav fotografije ali videoposnetki zlorabe nikoli niso bili najdeni ali ni bilo ugotovljeno, da izvirajo od zadevne žrtve. | The data above are subject to a number of additional caveats. The existing statistical data do not always allow Member States to separate data on victims who were identified based on a report from a provider from those where, for example, the victim himself or herself may have reported the criminal offence or someone else who knew the victim or detected the criminal offence may have reported it (as mentioned by Slovenia). Sweden reported that children who have been identified through chat logs are also part of the reporting although photos or videos of the abuse never have been found or determined to originate from that specific victim. |
Na splošno podatki v preglednici n.8 ne ustrezajo nujno obveznostim poročanja iz začasne uredbe, ki se nanašajo le na žrtve, ki so bile rešene zaradi prijav, ki so jih podali ponudniki in organizacije, ki delujejo v javnem interesu proti spolni zlorabi otrok na podlagi Uredbe. Podatki, predloženi v nekaterih primerih, vključujejo žrtve, identificirane iz različnih drugih razlogov in z drugimi sredstvi. | Overall, the data in the table n.8 do not necessarily correspond to the reporting obligations in the Interim Regulation, which refer only to victims rescued thanks to the reports submitted by providers and by organisations acting in the public interest against child sexual abuse under the Regulation. The data provided in some cases include victims identified due to a variety of other reasons and means. |
Podatki tako ne omogočajo celovitega pregleda števila otrok, identificiranih kot žrtve spolne zlorabe na spletu v EU. | The data thus do not enable a comprehensive overview on the number of children identified as victims of online child sexual abuse in the EU. |
Poleg tega, tudi če je bila žrtev identificirana, to ne pomeni nujno, da je prišlo do obsodbe, povezane z identifikacijo. V nekaterih primerih je bila žrtev identificirana, a v preiskavi ni bilo mogoče določiti osumljenca ali doseči obsodbe (Švedska). | In addition, even where a victim was identified, it does not necessarily mean that there has been a conviction linked to that identification. In some cases, the victim was identified, but the investigation could not establish a suspect or lead the investigation to a conviction (Sweden). |
Kljub temu je iz podatkov mogoče sklepati, da je bilo veliko število žrtev identificiranih zaradi prostovoljne prijave v skladu z začasno uredbo. To potrjujejo poročila organov kazenskega pregona o primerih, ki se pogosto sprožijo le na podlagi prostovoljne prijave 13 . | Nonetheless, it can be inferred from the data that a significant number of victims have been identified with the help of voluntary reporting in accordance with the Interim Regulation. This is confirmed by reports of cases from law enforcement authorities, which are often launched only on the basis of voluntary reporting 13 . |
2.2.3 Število obsojenih storilcev | 2.2.3. The number of perpetrators convicted |
Večina držav članic je izpolnila svoje obveznosti, dve državi članici pa nista predložili nobenih podatkov ali utemeljitve, zakaj nista poročali, v skladu s členom 8(1) (c) začasne uredbe. | While most Member States complied with their obligations, two Member States did not provide any data or justification for not reporting pursuant to Article 8(1)(c) of the Interim Regulation. |
Več držav članic ni predložilo nobenih statističnih podatkov za obdobje poročanja v skladu z navedeno določbo, navedle pa so naslednje razloge: | Several Member States did not provide any statistics for the reporting period pursuant to that provision and provided the following reasons: |
–podatki še niso na voljo (Belgija in Španija); | -data not available yet (Belgium and Spain); |
–centralna podatkovna zbirka, ki se uporablja za beleženje kaznivih dejanj, ni zahtevala beleženja vrste prvotne prijave (Irska); | -the central database used to record crime did not require to record the nature of the initial referral (Ireland); |
–podatki se ne zbirajo (Nemčija). | -the data is not collected (Germany). |
Države članice so sporočile zelo različne podatke o številu obsojenih storilcev, v zajetem obdobju poročanja pa ni bilo skladnosti, kot to prikazuje preglednica št. 9 spodaj. | Member States reported very varied data on the number of perpetrators convicted and there was no coherence in the reporting periods covered, as shown by table n.8 below. |
Preglednica št. 9: Število obsojenih storilcev | Table n.9: The number of perpetrators convicted |
Država | Obdobje poročanja | Število obsodb | Opombe | Country | Reporting period | Number of convictions | Comments |
Avstrija | 2021 | 850 | Podatki ne razlikujejo med kaznivimi dejanji, storjenimi na spletu in zunaj njega. | Austria | 2021 | 850 | The data do not differentiate between offences committed online and offline. |
Belgija | – | – | Podatki niso na voljo. | Belgium | - | - | Data not available. |
Bolgarija | 2021–2022 | 52 | Bulgaria | 2021 to 2022 | 52 |
Hrvaška | – | – | Podatki niso na voljo. | Croatia | - | - | Data not available. |
Ciper | 2022 | 0 | Še brez obsodb. | Cyprus | 2022 | 0 | No convictions so far. |
Češka | 1. januar 2022–31. julij 2022 | 20 | Czechia | 1 January 2022 to 31 July 2022 | 20 |
Danska | 2. avgust 2021–20. januar 2023 | 224 | Denmark | 2 August 2021 to 20 January 2023 | 224 |
Estonija | 2021 | 2 | Vključene so samo obsodbe na podlagi prijav centra NCMEC. | Estonia | 2021 | 2 | Includes only convictions resulting from NCMEC reports. |
Finska | 2021 | 240 | Finland | 2021 | 240 |
Francija | 4. avgust 2021–3. avgust 2022 | 820 | France | 4 August 2021 to 3 August 2022 | 820 |
Nemčija | – | – | Podatki niso na voljo. | Germany | - | - | Data not available. |
Grčija | 2021–2022 | 62 | Greece | 2021 to 2022 | 62 |
Madžarska | 2021–2022 | 126 | Hungary | 2021 to 2022 | 126 |
Irska | – | – | Podatki niso na voljo. | Ireland | - | - | Data not available. |
Italija | 2021–2022 | 5 835 | Podatki ne razlikujejo med kaznivimi dejanji, storjenimi na spletu in zunaj njega. | Italy | 2021 to 2022 | 5 835 | The data do not differentiate between offences committed online and offline. |
Latvija | 2021–2022 | 33 | Podatki ne razlikujejo med kaznivimi dejanji, storjenimi na spletu in zunaj njega. | Latvia | 2021 to 2022 | 33 | The data do not differentiate between offences committed online and offline. |
Litva | 1. januar 2021–30. junij 2022 | 10 | Lithuania | 1 January 2021 to 30 June 2022 | 10 |
Luksemburg | 2022 | 11 | Podatki ne razlikujejo med kaznivimi dejanji, storjenimi na spletu in zunaj njega. | Luxembourg | 2022 | 11 | The data do not differentiate between offences committed online and offline. |
Malta | – | – | Podatki niso bili predloženi/sporočeni. | Malta | - | - | Data not submitted/reported. |
Nizozemska | 2021 | 217 | Netherlands | 2021 | 217 |
Poljska | Od druge polovice leta 2021 do prve polovice leta 2022 / 2022 / prva polovica leta 2023 | 185 / 194 / 81 | Poland | Second half of 2021 to first half of 2022 / 2022 / First half of 2023 | 185 / 194 / 81 |
Portugalska | – | – | Podatki niso bili predloženi/sporočeni. | Portugal | - | - | Data not submitted/reported. |
Romunija | – | – | Podatki niso bili predloženi/sporočeni. | Romania | - | - | Data not submitted/reported. |
Slovaška | 2021 | 10 | Slovakia | 2021 | 10 |
Slovenija | 2021–2022 | 45 | Slovenia | 2021 to 2022 | 45 |
Španija | 2022 | – | Podatki niso na voljo. | Spain | 2022 | - | Data not available. |
Švedska | 2022 | 55 | Sweden | 2022 | 55 |
Opozoriti je treba, da se število obsodb ne ujema s številom obsojenih storilcev, saj je lahko oseba obsojena za eno ali več kaznivih dejanj spolne zlorabe otrok na spletu. | It is important to note that the number of convictions does not equal the number of perpetrators convicted, as a person might be convicted for one or more offences of child sexual abuse online. |
Poleg tega statistični podatki o obsodbah, sporočeni za določeno obdobje, niso nujno povezani s prijavami, ki so bile prejete v danem obdobju (tj. obsodba v letu 2022 bi lahko bila povezana s prijavo iz leta 2021 ali 2020, prijava iz leta 2022 pa lahko vodi do obsodbe v letu 2023 ali pozneje). To dejstvo je v svojih poročilih izrecno izpostavilo več držav članic (Irska, Luksemburg in Švedska). | What is more, the statistics on convictions reported for a certain period are not necessarily linked to the reports that have been received in that given period (i.e. a conviction in 2022 could for instance be linked to a report from 2021 or 2020 and a report from 2022 might only lead to a conviction in 2023 or later). This fact was explicitly highlighted by several Member States in their reports (Ireland, Luxembourg, Sweden). |
V nekaterih primerih se niso zbirali statistični podatki o tem, ali so prijave sumljivih dejavnosti (npr. prek centra NCMEC) privedle do obsodb, povedano drugače, da so bile obsodbe posledica informacij, ki jih je zagotovil ponudnik ali javna organizacija (Avstrija in Latvija). Samo Estonija je izrecno potrdila, da statistični podatki izkazujejo samo obsodbe, ki so posledica prijav centra NCMEC. Možno je tudi, da so prijave privedle so drugih storilcev, ki so predmet preiskav in so med preiskavami tudi obsojeni (Avstrija). | In some instances, no statistics were collected on whether reports of suspicious activities (e.g. via NCMEC) led to convictions, or in other words that convictions resulted from the information provided by a provider or a public organisation (Austria, Latvia). Only Estonia explicitly confirmed that the statistics show only convictions resulting from NCMEC reports. It is also possible that the reports led to other offenders, who are investigated and convicted in the course of the investigations (Austria). |
Večinoma se predvideva, da število sporočenih obsodb izhaja iz zaključka primera, potem ko je bila v pravosodnem sistemu domnevno vložena pritožba. V eni državi članici (Danska), kjer se podatki zbirajo glede na najnovejšo odločbo, podatki niso dokončni, saj so bile pozneje morda zoper odločbe vložene pritožbe. | Mostly it is presumed that the number of convictions reported is from when a case is finalised after having been presumably appealed through the judicial system. In one Member State (Denmark), where the numbers are compiled according to the latest decision, the numbers are not final, as the decisions may have been appealed afterwards. |
V nekaterih primerih podatki, ki so vključeni v nacionalne informacijske sisteme in jih države članice sporočajo, ne razlikujejo med kaznivimi dejanji, storjenimi na spletu ali zunaj njega (Avstrija, Luksemburg in Latvija). | In certain instances, the data contained in the national IT systems and reported by the Member States do not differentiate between offences committed online and offline (Austria, Luxembourg, Latvia). |
Zelo različna poročila, ki so jih predložile države članice, in način zbiranja statističnih podatkov na nacionalni ravni torej ne omogočajo celovitega pregleda števila storilcev, obsojenih zaradi spolne zlorabe otrok na spletu v EU. Trenutno – na podlagi razpoložljivih podatkov – tudi ni mogoče jasno povezati teh obsodb s prijavami, ki so jih podali ponudniki in organizacije, ki delujejo v javnem interesu proti spolni zlorabi otrok, v konkretnih obdobjih poročanja v skladu z Uredbo. | The very varied reports submitted by the Member States and the way statistical data are gathered at national level thus do not allow for a comprehensive overview of the number of perpetrators convicted for online child sexual abuse in the EU. It is also not possible at present – on the basis of the data available – to link these convictions clearly to reports submitted by providers and organisations acting in the public interest against child sexual abuse in concrete reporting periods in accordance with this Regulation. |
2.3Razvoj tehnološkega napredka | 2.3.Developments in technological progress |
Tehnologije, ki se zdaj uporabljajo za odkrivanje spolne zlorabe otrok na spletu, vključujejo tehnologije in orodja za odkrivanje „znanih“ (tj. že odkritih) posnetkov spolne zlorabe otrok, „novih“ (tj. prej še neodkritih) posnetkov spolne zlorabe otrok in pridobivanja otrok za spolne namene. | Technologies currently used to detect child sexual abuse online include technologies and tools to detect ‘known’ (i.e. previously detected) CSAM, ‘new’ (i.e. not previously detected) CSAM and solicitation of children (known as ‘grooming’). |
Spodaj navedeni primeri vključujejo nekaj najpogosteje uporabljenih orodij in ne predstavljajo izčrpnega seznama. Veliko teh orodij je na voljo ponudnikom, organom kazenskega pregona in drugim organizacijam, kadar je mogoče izkazati legitimen interes. Običajno se ta orodja uporabljajo skupaj s človeškim pregledom, da se zagotovi največja možna natančnost. | The examples given below include some of the most widely used tools, and they do not represent an exhaustive list. Many of these tools are made available to providers, law enforcement authorities and other organisations where a legitimate interest can be shown. Typically, these tools are combined with human review to ensure the maximum possible accuracy. |
V tem oddelku je vključen tudi dodaten razvoj tehnološkega napredka, povezanega z umetno inteligenco. | This section also includes additional developments in technological progress related to artificial intelligence. |
2.3.1 Odkrivanje znanih posnetkov spolne zlorabe otrok | 2.3.1. Known CSAM detection |
Obstoječe tehnologije za odkrivanje znanih posnetkov spolne zlorabe otrok se opirajo izključno na samodejno analizo vsebine 14 in običajno temeljijo na zgoščevanju. Tehnologija zgoščevanja je vrsta digitalnega prstnega odtisa. Ustvari edinstven digitalni podpis (znan kot „hash“) podobe, ki se nato primerja s podpisi („hashes“) drugih fotografij, da se najdejo kopije iste podobe. Ta tehnologija zaznava samo ujemajoče se edinstvene digitalne podpise in ne „vidi“ gradiva, ki se ne ujema z edinstvenim digitalnim podpisom. Zgoščene vrednosti tudi niso povratne in z njimi torej ni mogoče poustvariti podobe. | Existing technologies to detect known CSAM rely solely on automatic analysis of content 14 and are typically based on hashing. Hashing technology is a type of digital fingerprinting. It creates a unique digital signature (known as a “hash”) of an image which is then compared against signatures (hashes) of other photos to find copies of the same image This technology only detects matching hashes and does not 'see' any material which do not match the hash. Hash values are also not reversible, and therefore cannot be used to recreate an image. |
Obstaja veliko različic in izvedb tehnologije zgoščevanja. Orodja, opredeljena kot tista, ki se uporabljajo za odkrivanje znanih posnetkov spolne zlorabe otrok, vključujejo: (i) Microsoftov PhotoDNA ; (ii) Googlov CSAI Match ; (iii) Applov NeuralHash + Private Set Intersection ; (iv) Meta SSN++ ; (v) PDQ in TMK+PDQF ; (vi) MD5 Hash generator (Skype); (vii) Safer (Thorn). | Many variations and implementations of hashing technology exist. Tools identified as used for known CSAM detections include: (i) Microsoft PhotoDNA ; (ii) Google CSAI Match ; (iii) Apple NeuralHash + Private Set Intersection ; (iv) Meta SSN++ ; (v) PDQ and TMK+PDQF ; (vi) MD5 Hash generator (Skype); (vii) Safer (Thorn). |
Najpogosteje uporabljeno orodje je Microsoftov PhotoDNA, ki ga uporablja več kot 150 organizacij 15 . Orodje PhotoDNA se uporablja že več kot deset let in ima visoko stopnjo natančnosti. Stopnja lažno pozitivnih primerov je na podlagi testiranja ocenjena na ne več kot en primer na 50 milijard 16 . Stopnja napak orodja PhotoDNA zaradi narave tehnologije ostaja izredno nizka. Tehnologija odkriva izključno kopije že prepoznane vsebine. Prvotno orodje PhotoDNA odkriva znane posnetke spolne zlorabe otrok v podobah, na voljo pa je tudi različica za odkrivanje posnetkov spolne zlorabe otrok v videoposnetkih. | The most widely used tool is Microsoft PhotoDNA, used by over 150 organisations 15 . PhotoDNA has been in use for more than 10 years and has a high level of accuracy. The rate of false positives is estimated at no more than 1 in 50 billion, based on testing 16 . PhotoDNA’s error rate remains exceedingly low because of the nature of the technology. The technology exclusively detects copies of previously identified content. While the original PhotoDNA detects known CSAM in images, a version for detecting CSAM in videos is also available. |
Tehnologija se stalno razvija in izboljšuje. Microsoft je maja 2023 napovedal uvedbo novih zmogljivosti za iskanje ujemanj, ki omogočajo hitrejše iskanje (približno 350-krat hitrejše), hkrati pa znižujejo stroške postopka iskanja ujemanj brez izgube natančnosti. Po navedbah Microsofta nova knjižnica omogoča tudi celovitejše odkrivanje zrcalnih ali zasukanih slik. Poleg tega je organizacija Internet Watch Foundation (IWF) pred kratkim sporočila, da je izboljšala svojo tehnologijo zgoščevanja 17 . | The technology is continuously developing and being improved. In May 2023, Microsoft announced the deployment of new matching capabilities that enable swifter searching (around 350 times faster), while reducing the cost of the matching process with no loss of accuracy. According to Microsoft, the new library also enables more comprehensive detection of flipped or rotated images. In addition, the Internet Watch Foundation (IWF) reported recently enhancing its hashing technology 17 . |
2.3.2 Odkrivanje novih posnetkov spolne zlorabe otrok | 2.3.2. New CSAM detection |
Tehnologije, ki se zdaj uporabljajo za odkrivanje novih posnetkov spolne zlorabe otrok, vključujejo klasifikatorje in umetno inteligenco (UI), ki analizirajo podobe in videoposnetke za odkrivanje vzorcev vsebine, ki se ujemajo z vzorci, ustvarjenimi na podlagi že prepoznanih posnetkov spolne zlorabe otrok. Klasifikator je algoritem, ki lahko s prepoznavanjem vzorcev podatke razvrsti v označene razrede ali kategorije informacij. Klasifikatorji za učenje potrebujejo podatke, in več podatkov kot prejmejo, bolj natančni so. | Technologies currently used for the detection of new CSAM include classifiers and artificial intelligence (AI) that analyse images and videos to detect content patterns that match patterns generated on the basis of previously identified child sexual abuse materials. A classifier is an algorithm that sorts data into labelled classes, or categories of information, through pattern recognition. Classifiers need data to be trained on and their accuracy improves the more data they are fed. |
Orodja za odkrivanje novih posnetkov spolne zlorabe otrok vključujejo: (i) Safer (Thorn); (ii) Googlov Content Safety API ; (iii) Facebookova tehnologija umetne inteligence 18 ; (iv) Amazonov Rekognition ; (v) Hivova umetna inteligenca za vizualno vsebino . | Tools to detect new CSAM include: (i) Safer (Thorn); (ii) Google Content Safety API ; (iii) Facebook’s AI technology 18 ; (iv) Amazon Rekognition ; (v) Hive AI for visual content . |
Raziskave so pokazale, da so avtomatizirani orodja in sistemi, kot so klasifikatorji, najkoristnejša sredstva za odkrivanje posnetkov spolne zlorabe otrok 19 . Pri odkrivanju novih posnetkov spolne zlorabe otrok je stopnja natančnosti trenutno precej nad 90 %. Na primer, pri podjetju Thorn navajajo, da lahko svoj klasifikator za posnetke spolne zlorabe otrok nastavijo na 99-odstotno stopnjo natančnosti (za znane in nove posnetke spolne zlorabe otrok), kar pomeni, da stopnja lažno pozitivnih primerov znaša 0,1 % 20 . Z večjo uporabo in več povratnimi informacijami se bodo te metrike verjetno še izboljšale. | Research has shown that automated tools and systems such as classifiers are the most useful means of detecting CSAM. 19 For the detection of new CSAM, the accuracy rate currently lies significantly above 90%. For example, Thorn indicates that its CSAM Classifier can be set to 99% precision rate (for both known and new CSAM), meaning a 0,1% false positive rate 20 . These metrics are likely to improve with increased usage and feedback. |
2.3.3 Odkrivanje pridobivanja otrok za spolne namene | 2.3.3. Grooming detection |
Orodja za odkrivanje pridobivanja otrok za spolne namene v besedilnih komunikacijah izkoriščajo tehnologije izključno za odkrivanje vzorcev, ki nakazujejo možne konkretne elemente za domnevno spolno zlorabo otrok na spletu, ne da bi bila zmožna razbrati vsebino komunikacij. Tehnika se uporablja za besedilne klepete. Pogovori so ocenjeni na podlagi vrste značilnosti in dodeljena jim je skupna ocena verjetnosti, ki nakazuje ocenjeno verjetnost, da pogovor pomeni pridobivanje otrok za spolne namene. Te ocene so determinante, ki jih določijo posamezna podjetja, za označevanje pogovorov za dodatni človeški pregled. | Tools to detect grooming (solicitation of children) in text-based communications make use of technologies solely to detect patterns which point to possible concrete elements of suspicion of online child sexual abuse, without being able to deduce the substance of the content. The technique is applied to text-based chat conversations. Conversations are rated on a series of characteristics and assigned an overall probability rating, indicating the estimated probability that the conversation constitutes grooming. These ratings serve as a determinant, set by individual companies, to flag conversations for additional human review. |
Orodja, ki se uporabljajo za odkrivanje besedil, vključujejo: (i) Microsoftov Project Artemis 21 ; (ii) Amazonov Rekognition ; (iii) Twitchevo tehnologijo Spirit AI (na podlagi obdelave naravnega jezika, klasifikatorjev besedila) 22 ; (iv) Metin notranji interno izdelan klasifikator za „razvrščanje“ na podlagi strojnega učenja (ki interno tehnologijo za analizo jezika združuje z metapodatki); (v) filtriranje klepetov na platformi Roblox 23 ; (vi) tehnično rešitev podjetja Thorn in združenja Tech Coalition, ki temelji na strojnem učenju in klasifikatorjih 24 . | Tools used for text detection operations include: (i) Microsoft’s Project Artemis 21 ; (ii) Amazon Rekognition ; (iii) Twitch’s Spirit AI technology (based on NLP, text classifiers) 22 ; (iv) Meta in-house internally built machine learning ‘ranking’ classifier (combining internal language analysis tech with meta data); (v) Roblox chat filtering 23 ; (vi) Thorn and the Tech Coalition’s technical solution based on machine learning and classifiers 24 . |
Tako kot za prepoznavanje novih posnetkov spolne zlorabe otrok je tudi za prepoznavanje vsebine pridobivanja otrok za spolne namene treba tehnologijo podučiti s tako vsebino. Dostop do takih učnih podatkov ostaja največji izziv za razvoj in izboljšanje takih tehnologij. | As in the case of identification of new CSAM, identifying grooming content requires training the technology with such content. Access to such training data remains the biggest challenge to the development and improvement of such technologies. |
Podjetje Thorn je v partnerstvu z združenjem Tech Coalition in njegovimi člani sprožilo novo pobudo, namenjeno razvoju tehnične rešitve za odkrivanje in obravnavo poskusov pridobivanja otrok za spolne namene prek spleta, ki bo koristna in uporabna za vrsto platform, ki ponujajo besedilne komunikacije. Temeljila bo na ekipnem delu podjetja Thorn na področju razvoja klasifikatorja NLP (obdelava naravnega jezika) ali modela strojnega učenja, ki odkriva in razvršča, kadar spletna vsebina ali vedenje spada v opredeljene „razrede“, povezane s pridobivanjem otrok za spolne namene (na primer izpostavljenost gradivu s spolno vsebino ali prizadevanje za osebno srečanje z mladoletnikom), ter na splošni oceni, kako zelo je pogovor povezan s pridobivanjem otrok za spolne namene 25 . | Thorn, in partnership with the Tech Coalition and its members, has launched a new initiative aimed at developing technical solution to identify and address attempts of online grooming that will be useful and usable for a range of platforms offering text-based communications. It will be based on Thorn's team’s work on an NLP (natural language processing) classifier, or machine learning model, that detects and categorises when online content or behaviour falls into defined "classes" related to grooming (such as exposure to sexual material or seeking an in-person meetup with a minor) as well as an overall score for how related a conversation is to grooming. 25 |
2.3.4 Novi izzivi, ki jih prinašajo klepetalni boti z umetno inteligenco in generatorji umetnin/podob | 2.3.4. New challenges raised by Artificial intelligence chatbots and art/image generators |
Razvoj in izdaja klepetalnih botov z umetno inteligenco, kot je ChatGPT (velik jezikovni model (LLM), ki ga je razvila organizacija OpenAI ), in generatorjev umetnin/podob, kot sta DALL-E 26 in Midjourney 27 , sta deležna velike pozornosti javnosti, predvsem zaradi njihove sposobnosti hitrega zagotavljanja pripravljenih odgovorov ali ustvarjanja realističnih podob, ki jih je mogoče uporabiti v številnih različnih kontekstih. Ta nova orodja so hitro postala vsesplošno priljubljena in uporabljena. Tehnološka podjetja, vključno z Microsoftom in Googlom, financirajo in razvijajo vodilne produkte, nove tehnologije pa se izboljšujejo, tako da se redno uvajajo izboljšane različice. | The development and release of AI chatbots such as ChatGPT (a large language model (LLM) developed by OpenAI ) and art/image generators such as DALL-E 26 and Midjourney 27 , has generated significant public attention, mainly due to their ability to quickly provide ready-to-use answers or create realistic images that can be applied to a vast number of different contexts. These new tools rapidly gained widespread popularity and use. Leading products are being funded and developed by tech companies including Microsoft and Google and the new technologies are being refined and improved versions rolled out on a regular basis. |
Te tehnologije podjetjem in javnosti zagotavljajo velike priložnosti, vendar lahko predstavljajo tudi tveganja. Pomisleki v zvezi s takimi produkti vključujejo tudi vprašanje, kako jih bodo storilci kaznivih dejanj morda želeli izkoristiti za zlonamerno uporabo, vključno s spolnim izkoriščanjem otrok. | While these technologies offer great opportunities to businesses and the public alike, they can also pose a risk for them. Concerns about such products include how criminals may wish to exploit them for their nefarious purposes including child sexual exploitation. |
Kot poroča Europol, so vse informacije, ki jih zagotavlja ChatGPT, sicer prosto dostopne na internetu, vendar to orodje zlonamernim akterjem močno olajšuje, da se „brez predhodnega znanja podučijo o velikem številu potencialnih področij kriminala, od tega, kako vlomiti v stanovanje, do terorizma, kibernetske kriminalitete in spolne zlorabe otrok“. To navedenim osebam omogoča, da bolje razumejo te vrste kaznivih dejanj in jih posledično tudi izvedejo 28 . | As reported by Europol, while all the information ChatGPT provides is freely available on the internet, the tool makes it significantly easier for malicious actors “to learn about a vast number of potential crime areas with no prior knowledge, ranging from how to break into a home, to terrorism, cybercrime and child sexual abuse”. This enables said persons to better understand and subsequently carry out these types of crimes 28 . |
Pravila organizacije OpenAI omejujejo zmožnost orodja ChatGPT, da bi se odzvalo na pozive k spolni, sovražni ali nasilni vsebini ali vsebini, ki spodbuja samopoškodovanje. Kljub temu je mogoče te zaščitne ukrepe precej enostavno zaobiti s pisanjem pozivov 29 . Nedavna uvedba klepetalnih botov z umetno inteligenco (npr. v aplikaciji Snapchat) kaže, kako lahko ti klepetalni boti prestopijo mejo do žaljivih ali nevarnih interakcij, vključno s spolno zlorabo otrok 30 . Ker zdaj več podjetij razmišlja o testiranju klepetalnih botov z umetno inteligenco na svojih platformah (Instagram, morebiti WhatsApp in Messenger), je treba skrbno oceniti učinek na uporabnike, zlasti otroke in mlade. | OpenAI’s rules restrict ChatGPT’s capability to respond to prompts for sexual, hateful, violent content or content promoting self-harm. Nonetheless, these safeguards can be circumvented fairly easily through prompt engineering. 29 Recent deployment of AI chatbots (e.g. by Snapchat) shows how these can cross the line into offensive or dangerous interactions, including child sexual abuse 30 . With more companies now considering testing AI chatbots on their platforms (Instagram, potentially WhatsApp and Messenger), the impact on users, especially children and young people, has to be carefully assessed. |
Ta nova orodja zahtevajo tudi ustrezne zaščitne ukrepe, tako da se jih ne zlorablja za ustvarjanje globokih ponaredkov spolne zlorabe otrok z umetno inteligenco 31 . Glede na hitrost razvoja orodij umetne inteligence je verjetno, da bo kmalu postalo precej lažje ustvariti podobe, ki jih ne bo mogoče razlikovati od dejanskih podob. To pomeni več ključnih izzivov za boj proti spolni zlorabi otrok, saj bo lahko zmožnost organov kazenskega pregona za preiskovanje in kazensko preganjanje primerov posnetkov spolne zlorabe otrok in identifikacijo dejanskih žrtev močno ovirana, če bodo na spletu začeli močno prevladovati zelo realistični računalniško ustvarjeni posnetki spolne zlorabe otrok 32 . | These new tools also require adequate safeguards so that they are not misused to produce AI generated deep-fake child sexual abuse material. 31 With the pace of development of AI tools, it is likely that it will soon become significantly easier to generate images that are indistinguishable from actual images. This presents several key challenges in fighting CSA, as the ability for law enforcement to investigate and prosecute CSAM cases and identify real victims may become severely hindered if highly realistic computer-generated CSAM become highly prevalent online 32 . |
Raziskave kažejo, da je dostopanje do posnetkov spolne zlorabe otrok pogosto prvi korak k dejanski zlorabi, ne glede na to, ali posnetki prikazujejo resnično ali le realistično zlorabo in izkoriščanje 33 . Omejevanje razširjanja z umetno inteligenco ustvarjenih globokih ponaredkov spolne zlorabe otrok je torej ključno kot oblika preprečevanja na strani storilca. Še en ključni pomislek je, da lahko tisti, ki pridobivajo otroke za spolne namene, napredne zmogljivosti orodja ChatGPT za ustvarjanje besedila uporabljajo skupaj z obstoječo brezplačno umetno inteligenco, ki pretvori besedila v slike, ter hitro in enostavno ustvarijo vsebino za lažne profile in verodostojne pogovore z mladimi, da bi na spletu nagovarjali otroke. „Sam ChatGPT ljudi ne bo spodbujal k pridobivanju otrok za spolne namene prek spleta, vendar vsem omogoča, da s tehnologijo umetne inteligence izboljšajo pogovore, ki jih imajo z otroki na spletu, tako da postanejo prepričljivejši in se žrtvam zdijo verodostojnejši, kar pripomore k manipulaciji.“ 34 Generativna umetna inteligenca bi lahko prispevala k porastu števila primerov pridobivanja otrok za spolne namene prek spleta in celo k „obsežnejši avtomatizaciji pridobivanja otrok za spolne namene“ 35 . | Research has shown that accessing child sexual abuse material is often the first step towards hands-on abuse, regardless of whether the material depicts real or realistically looking abuse and exploitation 33 . Limiting the dissemination of AI generated deep-fake child sexual abuse material is therefore crucial as a form of offender-side prevention. Another key concern is that groomers can use the advanced text-generation powers of ChatGPT, combined with existing free text-to-image AI to generate quickly and easily content for fake profiles and plausible conversations with young people to target children online. “Whilst ChatGPT in itself won’t encourage people to become online groomers, it does allow anyone to feed the conversations they are having with children online through AI technology to make themselves more persuasive and credible to their victims, aiding manipulation.” 34 Generative AI could have the potential to contribute to a rise in online grooming cases and even to “automate child grooming at scale” 35 . |
3.SKLEPNE UGOTOVITVE | 3.CONCLUSIONS |
Izvedbeni ukrepi, ki so jih sprejeli ponudniki | Implementation measures taken by providers |
Iz poročil ponudnikov je razvidno, da so odkrivanje in prijavljanje spolne zlorabe otrok na spletu na podlagi začasne uredbe izvajali s pomočjo različnih tehnologij in postopkov za odkrivanje. Vsi ponudniki so poročali, da so prijave podali centru NCMEC. Kar zadeva vrsto in količine osebnih podatkov, ki so jih ponudniki obravnavali, so poročila pokazala, da so se zbirali različni podatki o prometu in da je bila raven podrobnosti obdelanih količin podatkov različna, kar Komisiji preprečuje, da bi pridobila enotne podatke na ravni EU v zvezi s ponudniki za zadevno obdobje poročanja (julij 2021–31. januar 2023). | Providers’ reporting showed that they have been carrying out detection and reporting of child sexual abuse online under the Interim Regulation using a variety of detection technologies and processes. All providers reported sending these reports to NCMEC. In terms of type and volumes of personal data handled by providers, the reporting showed an array of traffic data collected and a varied level of granularity of volumes of data processed which prevents the Commission from obtaining unified EU level data relating to providers for the reporting period in question (July 2021 to 31 January 2023). |
Ponudniki niso predložili podatkov o številu in deležih napak (lažno pozitivni primeri) različnih uporabljenih tehnologij, razčlenjenih po uporabljenih tehnologijah, kar pomeni, da uporabljajo večplastni pristop za odkrivanje spolne zlorabe otrok na spletu, ki je dopolnjen s človeškim pregledom. Hkrati so ponudniki vzpostavili širok nabor zaščitnih in drugih ukrepov, da bi pri svojem odkrivanju omejili in znižali stopnjo napak. Še več, ponudniki so poročali, da imajo vzpostavljene politike hrambe podatkov in zaščitne ukrepe za varstvo podatkov, ki so opredeljene v njihovih politikah ali izjavah o varstvu osebnih podatkov ter podprte z zaščitnimi in drugimi ukrepi za varstvo podatkov, standardnimi za panogo. | Providers did not submit the number and ratios of errors (false positives) of the different technologies used broken down by technology employed, indicating that they use a layered approach to detection of online CSA complemented by human review. At the same time, providers put in place a wide range of measures and safeguards to limit and reduce the error rate in their detection. What is more, providers reported having data retention policies and data protection safeguards in place, defined in their Privacy Policies or Statements and supported by industry standard data protection safeguards and measures. |
Izvedbeni ukrepi, ki so jih sprejele države članice | Implementation measures taken by Member States |
Države članice morajo v skladu z začasno uredbo (njenim členom 8) predložiti tudi ključne statistične podatke o primerih spolne zlorabe otrok na spletu, ki jih odkrijejo in prijavijo organom kazenskega pregona, o številu identificiranih otrok, ki so žrtve, in o številu obsojenih storilcev. Ker so države članice večinoma predložile podatke za različna obdobja poročanja, iz predloženih podatkov ni bilo mogoče izračunati skupnega števila prijav odkrite spolne zlorabe otrok na spletu, prejetih na ravni EU. Poleg tega se lahko prijave, ki so jih države članice prejele in o njih poročale, razlikujejo od prijav, na podlagi katerih je mogoče ukrepati, tj. od prijav, ki jih je mogoče uporabiti za preiskave, ali od števila prijavljenih primerov. Le nekaj držav članic je navedlo vrsto ponudnikov, pri storitvah katerih je bila odkrita spolna zloraba otrok na spletu. V nekaterih primerih nacionalni statistični podatki ne razlikujejo med kaznivimi dejanji, preiskanimi na podlagi prijav ponudnikov in drugih organizacij, ki delujejo v javnem interesu proti spolni zlorabi otrok, in kaznivimi dejanji, preiskanimi na podlagi drugih prijav. | The Interim Regulation also obliges Member States (pursuant to its Article 8) to provide key statistics on cases of online child sexual abuse detected and reported to their law enforcement authorities, the number of children victims identified, and the number of perpetrators convicted. As Member States mostly provided data for differing reporting periods, it was not possible to calculate from the data submitted the total number of reports of detected online child sexual abuse received at EU level. In addition, the reports received and reported by Member States might differ from actionable reports, i.e. from reports that could be used for investigations, or number of cases reported. Only a few Member States indicated the type of providers on whose services the online child sexual abuse was detected. In some cases, the national statistical data do not differentiate between offences investigated on the basis of reports submitted by providers, and other organisations acting in the public interest against child sexual abuse and offences investigated on the basis of other reports. |
Tako iz prejetih poročil ni bilo mogoče pridobiti podatka o skupnem številu otrok, ki so bili v EU identificirani kot žrtve spolne zlorabe na spletu, glede na spol. Razlogi na primer vključujejo: sporočanje podatkov za različna obdobja; različne starostne omejitve za opredelitev otrok, ki so žrtve spolne zlorabe na spletu; nezbiranje tako podrobnih statističnih podatkov na nacionalni ravni zaradi tehničnih ali drugih omejitev; nerazlikovanje med otroki, ki so žrtve spolne zlorabe na spletu in zunaj njega, itd. Nekaj držav članic v statistične podatke vključuje tudi otroke, ki so sami ustvarili posnetke. Kar je še pomembneje, statistični podatki pogosto ne razlikujejo med žrtvami, identificiranimi na podlagi prijav ponudnikov in organizacij, ki delujejo v javnem interesu proti spolni zlorabi otrok na podlagi Uredbe, in žrtvami, ki so bile identificirane na podlagi drugih razlogov in sredstev. | Correspondingly, it was not possible to extract from the reports received the total number of children identified as victims of online child sexual abuse in the EU, differentiated by gender. The reasons include, for example: reporting data for differing periods; different age limits used to define child victims of CSA online; non-collection of statistics at such level of detail at the national level due to technical or other limitations; no differentiation between child victims of CSA online and offline etc. Some Member States include in their statistics also children that produced self-generated material. More importantly, the statistics often do not differentiate between victims identified based on reports submitted by providers and organisation acting in the public interest against child sexual abuse under the Regulation and those identified based on other reasons and means. |
Tudi pregled števila obsojenih storilcev je razdrobljen. V določenih primerih taki podatki niso na voljo, saj v podatkovnih zbirkah ni zabeležen vir prvotne prijave in zato podatki ne razlikujejo med storilci, obsojenimi na podlagi prijav, podanih na podlagi Uredbe, in na podlagi drugih prijav. V nekaterih primerih podatki, vsebovani v nacionalnih informacijskih sistemih, prav tako ne razlikujejo med kaznivimi dejanji, storjenimi na spletu in zunaj njega. Še več, statistični podatki o obsodbah, sporočeni za določeno obdobje, niso nujno povezani s prijavami, prejetimi v zadevnem obdobju, ampak se lahko nanašajo na prijave iz prejšnjih obdobij. Zbrani statistični podatki o številu obsodb se prav tako lahko razlikujejo od podatkov o številu obsojenih storilcev (en storilec je lahko večkrat obsojen). | The overview of the numbers of perpetrators convicted is also fragmented. In certain cases, such data are not available as the databases did not record the source of the initial referral and thus the data do not differentiate between perpetrators convicted because of reports submitted pursuant to the Regulation and other reports. In some instances, the data contained in national IT systems also do not differentiate between offences committed online and offline. What is more, the statistics on convictions reported for a certain period are not necessarily linked to the reports that have been received in that given period but might relate to reports from earlier periods. Collected statistics on the number of convictions might also differ from the number of perpetrators convicted (as one perpetrator might have more convictions). |
Raznoliki statistični podatki, ki so jih predložile države članice, za katere se zdi, da podatkov ne zbirajo vedno sistematično in ustrezno, in vsi navedeni dejavniki torej ne omogočajo celovitega pregleda prejetih prijav spolne zlorabe otrok na spletu, števila otrok, identificiranih kot žrtve tega kaznivega dejanja, ali števila storilcev, obsojenih na ravni EU na podlagi Uredbe. Dejstvo, da se za večino držav članic zdi, da obstajajo velike razlike med številom prijav, za katere center NCMEC navaja, da jih je poslal državam članicam, in številom prijav, za katere države članice navajajo, da so jih prejele, kaže, da zbiranje in sporočanje podatkov nista popolna. Nekatere države članice so potrdile, da so njihovi pristojni organi v postopku strukturnih sprememb ali reorganizacije, povezane z vzpostavljanjem novih oddelkov, pristojnih za preiskovanje kaznivih dejanj, povezanih s spolno zlorabo otrok na spletu. Vzpostavljajo se tudi novi informacijski sistemi, nacionalni organi v nekaterih državah članicah pa so bili pozvani, naj spremenijo svoje postopke registracije in statistične podatke. To bi moralo ustvariti ugodnejše pogoje za natančnejše statistične podatke držav članic v prihodnje. Komisija bo v vsakem primeru uporabila svoja pooblastila iz Pogodb za zagotovitev, da bodo države članice izpolnjevale svoje obveznosti poročanja iz začasne uredbe. | The heterogenous statistics submitted by the Member States, which appear to not always systematically and properly collect the data, and all the above-mentioned factors thus do not enable a comprehensive overview on the reports on online CSA received, the number of children identified as victims of this crime, or the number of perpetrators convicted at EU level pursuant to the Regulation. The fact that for most Member States there appears to be a significant disparity between the number of reports NCMEC lists as having sent to the Member State, and the number of reports the Member State lists as received, suggests that the data collection and reporting is not complete. Some Member States confirmed that their competent authorities are undergoing structural changes or reorganisation linked to the creation of new departments responsible for investigating crimes related to child sexual abuse online. New IT systems are also being put in place and national authorities were asked to alter their registration procedures and statistics in some Member States. This should create favourable conditions for having more accurate statistics from Member States in the future. In any case, the Commission will make use of its powers under the Treaties as needed to ensure that Member States comply with their reporting obligations under the Interim Regulation. |
Splošne ugotovitve | General considerations |
Na splošno to poročilo kaže, da ponudniki in države članice zelo različno sporočajo podatke v zvezi z bojem proti spolni zlorabi otrok na spletu na podlagi začasne uredbe. Večja standardizacija razpoložljivih podatkov, na primer tistih, navedenih v predlogu uredbe o preprečevanju spolne zlorabe otrok in boju proti njej 36 , in njihovo sporočanje bi prispevala k pridobitvi boljšega pregleda nad zadevnimi dejavnostmi na področju boja proti temu kaznivemu dejanju. Zdi se, da so potrebna dodatna prizadevanja ponudnikov in držav članic, da se zagotovita zbiranje podatkov in poročanje v skladu z zahtevami začasne uredbe. | Overall, this report shows considerable disparities in the reporting on data relating to combating online CSA under the Interim Regulation by both providers and Member States. Greater standardisation of available data and the reporting thereof, such as those provided in the proposal for a Regulation to prevent and combat child sexual abuse 36 , would contribute to obtaining a better picture as regards relevant activities in the fight against this crime. It appears that additional efforts by providers and Member States are needed to ensure data collection and reporting in line with the requirements of the Interim Regulation. |
Razpoložljivi podatki kažejo, da je v okviru sedanjega sistema prostovoljnega odkrivanja in prijavljanja mogoče, da se za posnetke, ki so samodejno označeni kot možni posnetki spolne zlorabe otrok, po človeškem pregledu izkaže, da to niso. To je lahko posledica neobstoja skupnega nabora edinstvenih digitalnih podpisov in drugih kazalnikov za odkrivanje posnetkov spolne zlorabe otrok, ki so potrjeno nezakoniti v EU, ali različnih pravnih standardov v jurisdikcijah, predvsem med EU in ZDA, zlasti glede ustreznih opredelitev. Podatki izkazujejo tudi velike razlike v številu zahtevkov za pregled in stopnjah uspešnosti pregleda, iz katerih ni mogoče oblikovati sklepov, saj ni dovolj informacij zlasti o obsegu zahtevkov za pregled in razlogih za obnovitev. | The available data shows that under the current voluntary detection and reporting system it is possible that materials automatically flagged as possible CSAM result, upon human review, to be not CSAM. This can be due to the lack of a common set of hashes and other indicators to detect CSAM, confirmed as illegal in the EU, or differing legal standards across jurisdictions, notably between the EU and the US, in particular on the relevant definitions. The data also suggests large variations in the number of review requests, and review success rates, from which it is not possible to extract conclusions, given the lack of information on notably the scope of the review requests and the reasons to reinstate. |
Kar zadeva zahteve iz člena 9(2) o pogojih za obdelavo, predložene informacije kažejo, da uporabljene tehnologije ustrezajo tehnološkim aplikacijam, zasnovanim izključno za odkrivanje in odstranjevanje posnetkov spolne zlorabe otrok na spletu ter njihovo prijavljanje organom kazenskega pregona in organizacijam, ki delujejo v javnem interesu proti spolni zlorabi otrok. V zvezi s tem, ali je bila uvedba tehnologij v skladu s stanjem tehnike in izvedena na način, ki čim manj posega v zasebnost, ter v zvezi s tem, ali sta bila izvedena predhodna ocena učinka v zvezi z varstvom podatkov iz člena 35 Uredbe (EU) 2016/679 in postopek predhodnega posvetovanja iz člena 36 navedene uredbe, ni bilo predloženih informacij. | As regards the requirements of Article 9(2) on the conditions for the processing, the information provided indicates that the technologies used correspond to technological applications designed for the sole purpose of detecting and removing online child sexual abuse material and reporting it to law enforcement authorities and to organisations acting in the public interest against child sexual abuse. No information was provided in relation to whether the deployment of the technologies was in accordance with the state of the art and in the least privacy-intrusive way, and on whether a prior data protection impact assessment as referred to in Article 35 of Regulation (EU) 2016/679 and a prior consultation procedure as referred to in Article 36 of that Regulation had been conducted. |
Kar zadeva sorazmernost Uredbe, je osrednje vprašanje, ali začasna uredba dosega ravnovesje, ki si ga je zakonodajalec Unije prizadeval doseči med, na eni strani, doseganjem cilja splošnega interesa v zvezi z učinkovitim bojem proti zadevnim zelo hudim kaznivim dejanjem in potrebo po varstvu temeljnih pravic otrok (dostojanstvo, osebna celovitost, prepoved nečloveškega ali ponižujočega ravnanja, zasebno življenje, pravice otroka) ter, na drugi strani, varstvom temeljnih pravic uporabnikov zajetih storitev (zasebnost, varstvo osebnih podatkov, svoboda izražanja, učinkovita pravna sredstva). Razpoložljivi podatki ne zadoščajo za oblikovanje dokončnih sklepov v zvezi s tem. Številčnega standarda ni mogoče in tudi ne bi bilo ustrezno uporabiti za oceno take sorazmernosti z vidika števila rešenih otrok, saj spolna zloraba močno negativno vpliva na otrokovo življenje in pravice. Kljub temu na podlagi navedenega nič ne kaže na to, da odstopanje ni sorazmerno. | As regards the proportionality of the Regulation, the central question is whether the Interim Regulation achieves the balance that the EU legislature sought to strike between, on the one hand, achieving the general interest objective of effectively combating the extremely serious crimes at issue and the need to protect the fundamental rights of children (dignity, integrity, prohibition of inhuman or degrading treatment, private life, rights of the child) and, on the other hand, safeguarding the fundamental rights of the users of the services covered (privacy, personal data protection, freedom of expression, effective remedy). The available data is insufficient to draw definitive conclusions in this respect. It is not possible nor would it be appropriate to apply a numerical standard when assessing such proportionality in terms of number of children rescued, given the significant negative impact on a child’s life and rights that is caused by sexual abuse. Nonetheless, in light of the foregoing, there are no indications that the derogation is not proportionate. |
Kljub pomanjkljivostim razpoložljivih podatkov, ki ne omogočajo vpogleda v uporabo prostovoljnih prijav v večjem številu držav članic, je iz podatkov, ki so na voljo, jasno, da je bilo v obdobju poročanja identificiranih več tisoč otrok, da je bilo doseženih več kot 2 000 obsodb in da je bilo iz obtoka odstranjenih več milijonov podob in videoposnetkov, zaradi česar se je zmanjšala sekundarna viktimizacija. Zato je mogoče skleniti, da je prostovoljno prijavljanje pomembno prispevalo k zaščiti velikega števila otrok, tudi pred stalno zlorabo, pri čemer se zdi, da je začasna uredba učinkovita. | Despite the shortcomings of the available data, which do not allow insight into the use of voluntary reports in a significant number of Member States, it is clear from the data that is available that thousands of children were identified in the reporting period, more than two thousand convictions were obtained, and millions of images and videos were removed from circulation, reducing secondary victimisation. Therefore, it can be concluded that voluntary reporting contributed significantly to the protection of a large number of children, including from ongoing abuse, and it appears that the Interim Regulation is effective. |
(1) | (1) |
Podjetje Twitter je poročilo predložilo pred preimenovanjem; v tem poročilu se navaja kot podjetje „X“. | Twitter submitted their contribution before its renaming; it will be referred to as “X” throughout the rest of this report. |
(2) | (2) |
Meta in X nista izrecno navedla konkretnega člena. | Meta and X did not specify the concrete Article explicitly. |
(3) | (3) |
Meta in X nista izrecno navedla konkretnega člena. | Meta and X did not specify the concrete Article explicitly. |
(4) | (4) |
Ponudniki, ki so navedeni v oklepajih, so izrecno poročali o konkretnih ukrepih. Če nekateri ponudniki niso navedeni, to ne pomeni, da ne izvajajo tega ukrepa, ampak le, da ga v svojem poročilu niso navedli. | The providers indicated in brackets are those that specifically reported the concrete measures. If some providers are not listed, it does not mean that they do not implement this measure, but only that they have not mentioned it in their report. |
(5) | (5) |
Politika zasebnosti podjetja X, 4. How Long We Keep Information (Kako dolgo hranimo informacije?), na voljo na spletnem naslovu: https://twitter.com/en/privacy. | X Privacy Policy, 4. How Long We Keep Information, Available at: https://twitter.com/en/privacy |
(6) | (6) |
Ponudniki, ki so navedeni v oklepajih, so izrecno poročali o konkretnih ukrepih. Če nekateri ponudniki niso navedeni, to ne pomeni, da ne izvajajo tega ukrepa, ampak le, da ga v svojem poročilu niso navedli. | The providers indicated in brackets are those that specifically reported the concrete measures. If some providers are not listed, it does not mean that they do not implement this measure, but only that they have not mentioned it in their report. |
(7) | (7) |
Informacije o organizacijah, ki delujejo v javnem interesu in katerim ponudniki prijavijo spolno zlorabo otrok na spletu na podlagi Uredbe, so objavljene na spletnem naslovu https://home-affairs.ec.europa.eu/policies/internal-security/child-sexual-abuse/legal-framework-protect-children_en v skladu z obveznostmi Komisije na podlagi člena 8(2) začasne uredbe. | The information on the organisations acting in the public interest to which providers report online child sexual abuse under this Regulation has been published at https://home-affairs.ec.europa.eu/policies/internal-security/child-sexual-abuse/legal-framework-protect-children_en , in line with the Commission’s obligations under Article 8(2) of the Interim Regulation. |
(8) | (8) |
Poročilo, ki ga je Nemčija predložila v skladu s členom 8 Uredbe (EU) 2021/1232, prejeto 18. oktobra 2022. Center NCMEC v svojih poročilih po državah v sistemu CyberTipline objavlja vse podatke o prejetih prijavah, ki se nanašajo na države članice EU, vključno z Nemčijo. Glej: Center NCMEC, 2021 CyberTipline Reports by Country (Poročila CyberTipline po državah za leto 2021), obiskano julija 2023; center NCMEC, 2022 CyberTipline Reports by Country (Poročila CyberTipline po državah za leto 2022), obiskano julija 2023. | Report submitted by Germany in line with article 8 of the Regulation (EU) 2021/1232, received on 18. October 2022. NCMEC publishes all data on reports received and related to the EU Member States, including Germany, in their CyberTipline Reports by Country. See: NCMEC, 2021 CyberTipline , accessed in July 2023; NCMEC, 2022 CyberTipline Reports by Country, accessed in July 2023. |
(9) | (9) |
https://www.bka.de/SharedDocs/Kurzmeldungen/DE/Kurzmeldungen/230623_Mindestspeicherfristen_IP-Adressen.html | https://www.bka.de/SharedDocs/Kurzmeldungen/DE/Kurzmeldungen/230623_Mindestspeicherfristen_IP-Adressen.html |
(10) | (10) |
Center NCMEC, 2021 CyberTipline Reports by Country (Poročila CyberTipline po državah za leto 2021), obiskano novembra 2023. | NCMEC, 2021 CyberTipline , accessed in November 2023. |
(11) | (11) |
Center NCMEC, 2022 CyberTipline Reports by Country (Poročila CyberTipline po državah za leto 2022), obiskano novembra 2023. | NCMEC, 2022 CyberTipline Reports by Country, accessed in November 2023. |
(12) | (12) |
Podatki centra NCMEC so na voljo tukaj . | NCMEC data is available here . |
(13) | (13) |
Glej na primer seznam vzorčnih primerov v EU, ki so bili sproženi zaradi prostovoljne prijave podjetij, v oceni učinka , priloženi predlogu uredbe o preprečevanju spolne zlorabe otrok in boj proti njej (glej zlasti Prilogo 7). | See for example a list of sample cases across the EU that were launched thanks to voluntary reporting by the companies in the impact assessment accompanying the Proposal for a Regulation to prevent and combat child sexual abuse (see in particular Annex 7). |
(14) | (14) |
Ponudniki menijo, da metapodatki niso učinkovito orodje za odkrivanje posnetkov spolne zlorabe otrok. Glej zlasti strani 10 in 11 dokumenta Content-Oblivious Trust and Safety Techniques: Results from a Survey of Online Service Providers (Tehnike zaupanja in varnosti, neodvisne od vsebine: rezultati ankete za ponudnike spletnih storitev), Pfefferkorn, R., Stanford Internet Observatory, 9, september 2021. | Providers do not consider metadata as an effective tool in detecting CSAM. See in particular p.10-11 of Pfefferkorn, R., Stanford Internet Observatory, Content-Oblivious Trust and Safety Techniques: Results from a Survey of Online Service Providers, 9 September, 2021. |
(15) | (15) |
Microsoft, Digital Crimes Unit (Enota za kibernetsko kriminaliteto). | Microsoft, Digital Crimes Unit . |
(16) | (16) |
Pričanje Hanyja Farida, razvijalca orodja PhotoDNA, pred odborom predstavniškega doma ZDA za energijo in trgovino, Fostering a Healthier Internet to Protect Consumers (Spodbujanje bolj zdravega interneta za zaščito potrošnikov), 16. oktober 2019. | Testimony of Hany Farid, PhotoDNA developer, to House Committee on Energy and Commerce Fostering a Healthier Internet to Protect Consumers, 16 October 2019. |
(17) | (17) |
Internet Watch Foundation (IWF), The Annual Report 2022 (Letno poročilo 2022), str. 129–133. | Internet Watch Foundation (IWF), The Annual Report 2022 , p.129–133. |
(18) | (18) |
Glej tukaj in tukaj za več informacij o Facebookovem orodju za proaktivno odkrivanje otroške golote in še neznane vsebine izkoriščanja otrok s pomočjo umetne inteligence in strojnega učenja. | See here and here for more information on Facebook’s tool to proactively detect child nudity and previously unknown child exploitative content using artificial intelligence and machine learning. |
(19) | (19) |
Pfefferkorn, R.: Content-Oblivious Trust and Safety Techniques: Results from a Survey of Online Service Providers (Tehnike zaupanja in varnosti, neodvisne od vsebine: rezultati ankete za ponudnike spletnih storitev), Journal of Online Trust and Safety, februar 2022, str. 1–38. | Pfefferkorn, R.: Content-Oblivious Trust and Safety Techniques: Results from a Survey of Online Service Providers, Journal of Online Trust and Safety, February 2022, p. 1-38. |
(20) | (20) |
Thorn, Thorn’s Automated Tool to Remove Child Abuse Content at Scale Expands to More Platforms through AWS Marketplace (Thornovo avtomatizirano orodje za odstranjevanje vsebine spolne zlorabe otrok v ustreznem obsegu se širi na več platform prek tržnice AWS Marketplace), 24. maj 2021. | Thorn, Thorn’s Automated Tool to Remove Child Abuse Content at Scale Expands to More Platforms through AWS Marketplace , 24 May 2021. |
(21) | (21) |
Microsoftov Project Artemis je bil razvit v sodelovanju s podjetji The Meet Group, Roblox, Kik in Thorn. | Microsoft’s Project Artemis was developed in collaboration with The Meet Group, Roblox, Kik and Thorn. |
(22) | (22) |
Za več informacij glej: https://safety.twitch.tv/s/article/Our-Work-to-Combat-Online-Grooming?language=en_US . | For more information see: https://safety.twitch.tv/s/article/Our-Work-to-Combat-Online-Grooming?language=en_US |
(23) | (23) |
Platforma Roblox filtrira objave in klepete igralcev, starih 12 let in mlajših, za odkrivanje neprimerne vsebine in za preprečevanje objave osebnih podatkov, npr. domačih naslovov. Ta sistem filtriranje zajema vsa področja komunikacije na platformi Roblox, javne in zasebne. Roblox, Safety Features: Chat, Privacy & Filtering (Varnostni elementi: klepet, zasebnost in filtriranje), obiskano julija 2023. | Roblox filters posts and chats for players age 12 and younger for inappropriate content and to prevent personal information from being posted, e.g. home addresses. This filtering system covers all areas of communication on Roblox, public and private. Roblox, Safety Features: Chat, Privacy & Filtering , accessed in July 2023. |
(24) | (24) |
Tech Coalition, New Technology to Help Companies Keep Young People Safe (Nova tehnologija, ki podjetjem pomaga zaščititi mlade), 20. junij 2023. | Tech Coalition, New Technology to Help Companies Keep Young People Safe , 20 June 2023. |
(25) | (25) |
Tech Coalition, New Technology to Help Companies Keep Young People Safe (Nova tehnologija, ki podjetjem pomaga zaščititi mlade), 20. junij 2023. | Tech Coalition, New Technology to Help Companies Keep Young People Safe , 20 June 2023. |
(26) | (26) |
DALL-E je sistem umetne inteligence, ki lahko ustvari realistične podobe in umetnine na podlagi opisa v naravnem jeziku. | DALL-E is an AI system that can create realistic images and art from a description in natural language. |
(27) | (27) |
Midjourney je generativni program, ki temelji na umetni inteligenci, storitev pa ustvarja podobe na podlagi opisov v naravnem jeziku. | Midjourney is a generative artificial intelligence program and service generates images from natural language descriptions. |
(28) | (28) |
Europol, ChatGPT - The impact of Large Language Models on Law Enforcement (ChatGPT – vpliv velikih jezikovnih modelov na kazenski pregon), 2023, ISBN 978-92-95220-57-7, str. 7. | Europol, ChatGPT - The impact of Large Language Models on Law Enforcement , 2023, ISBN 978-92-95220-57-7, page 7. |
(29) | (29) |
Swanson, S. M., ChatGPT Generated Child Sex Abuse When Asked to Write BDSM Scenarios (Spolna zloraba otrok, ki jo je ustvaril ChatGPT kot odgovor na poziv, naj napiše scenarije za BDSM), Vice, 6. marec 2023; Mitchell, A., ChatGPT gives sick child sex abuse answer, breaking its rules (ChatGPT da sprevržen odgovor o spolni zlorabi otrok in prekrši lastna pravila), New York Post, 24. julij 2023; Europol, ChatGPT - The impact of Large Language Models on Law Enforcement (ChatGPT – vpliv velikih jezikovnih modelov na kazenski pregon), 2023, ISBN 978-92-95220-57-7, str. 5. | Swanson, S. M., ChatGPT Generated Child Sex Abuse When Asked to Write BDSM Scenarios , Vice, 6 March 2023; Mitchell, A., ChatGPT gives sick child sex abuse answer, breaking its rules, New York Post, 24 July 2023; Europol, ChatGPT - The impact of Large Language Models on Law Enforcement , 2023, ISBN 978-92-95220-57-7, page 5. |
(30) | (30) |
Fowler, G.A., Snapchat tried to make a safe AI. It chats with me about booze and sex (Snapchat je poskusil ustvariti varno umetno inteligenco. Z mano je klepetal o alkoholu in seksu), The Washington Post, 14. marec 2023; Vincent, J., Instagram is apparently testing an AI chatbot that lets you choose from 30 personalities (Instagram očitno testira klepetalni bot z umetno inteligenco, ki omogoča izbiranje med 30 osebnostmi), The Verge, 7. julij 2023. | Fowler, G.A., Snapchat tried to make a safe AI. It chats with me about booze and sex , The Washington Post, 14 March 2023; Vincent, J., Instagram is apparently testing an AI chatbot that lets you choose from 30 personalities , The Verge, 7 July 2023. |
(31) | (31) |
Crawford, A., Smith, T., Illegal trade in AI child sex abuse images exposed (Razkrita nezakonita trgovina s podobami spolne zlorabe otrok, ustvarjenimi z umetno inteligenco), BBC, 27. junij 2023. | Crawford, A., Smith, T., Illegal trade in AI child sex abuse images exposed , BBC, 27 June 2023. |
(32) | (32) |
Thiel, D., Stroebel, M., in Portnoff, R. (2023). Generative ML and CSAM: Implications and Mitigations (Generativno strojno učenje in posnetki spolne zlorabe otrok: posledice in preprečevanje). Stanford Digital Repository. Na voljo na spletnem naslovu https://purl.stanford.edu/jv206yg3793 . https://doi.org/10.25740/jv206yg3793 . Str. 2. | Thiel, D., Stroebel, M., and Portnoff, R. (2023). Generative ML and CSAM: Implications and Mitigations . Stanford Digital Repository. Available at https://purl.stanford.edu/jv206yg3793 . https://doi.org/10.25740/jv206yg3793 . P. 2. |
(33) | (33) |
Protect Children, Protect Children’s research in the dark web is revealing unprecedented data on CSAM users (Raziskava temnega spleta organizacije Protect Children razkriva podatke brez primere o uporabnikih posnetkov spolne zlorabe otrok), 6. junij 2021; RAINN, What is Child Sexual Abuse Material (CSAM) (Kaj so posnetki spolne zlorabe otrok?), 25. avgust 2022. | Protect Children, Protect Children’s research in the dark web is revealing unprecedented data on CSAM users , 6 June 2021; RAINN, What is Child Sexual Abuse Material (CSAM) , 25 August 2022. |
(34) | (34) |
Breck Foundation, Is artificial intelligence putting children at risk? (Ali umetna inteligenca ogroža otroke?), 9. februar 2023, posodobljeno 3. aprila 2023. | Breck Foundation, Is artificial intelligence putting children at risk? , 9 February 2023, updated 3 April 2023. |
(35) | (35) |
Butler, J., AI tools could be used by predators to ‘automate child grooming’, eSafety commissioner warns (Komisarka za e-varnost opozarja, da bi lahko predatorji orodja umetne inteligence izkoriščali za „avtomatizacijo pridobivanja otrok za spolne namene“), The Guardian, 19. maj 2023 . | Butler, J., AI tools could be used by predators to ‘automate child grooming’, eSafety commissioner warns , The Guardian, 19 May 2023. |
(36) | (36) |
Predlog uredbe Evropskega parlamenta in Sveta o določitvi pravil za preprečevanje spolne zlorabe otrok in boj proti njej, COM(2022) 209 final . | Proposal for a Regulation of the European Parliament and of the Council laying down rules to prevent and combat child sexual abuse, COM/2022/209 final . |