COM(2022) 414 final
Recommendation for a
authorising the opening of negotiations on behalf of the European Union for a Council of Europe convention on artificial intelligence, human rights, democracy and the rule of law
CONTEXT OF THE PROPOSAL
Artificial intelligence (AI) offers great opportunities for society, the environment and the economy. At the same time, depending on the circumstances regarding its specific application and use, certain AI may cause material or immaterial harm and may generate risks to public interests and individuals’ fundamental rights and freedoms. A coherent legal framework for AI, protecting such public interests and fundamental rights, while fostering trust and innovation, is therefore crucial to addressing those challenges and making use of AI’s potential.
In April 2021, the European Commission proposed a comprehensive regulation on AI that would harmonise the rules for AI systems in all 27 EU Member States. The European Parliament and the Council are currently negotiating the proposal in accordance with the ordinary legislative procedure. Various international organisations, including the Council of Europe, have also stepped up their efforts in this field, recognising the cross-border nature of AI and the need for international cooperation to address the common challenges.
Towards a Council of Europe convention on artificial intelligence, human rights, democracy and the rule of law
On 11 September 2019, the Council of Europe Committee of Ministers decided to set up an Ad hoc Committee on Artificial Intelligence (CAHAI). Its task was to examine the feasibility and potential elements of a legal framework on the development, design and application of AI, while taking into account the Council of Europe’s standards on human rights, democracy and the rule of law as well as relevant existing international legal instruments. CAHAI’s work was based on a feasibility study and a multi-stakeholder consultation carried out in the spring of 2021. On 3 December 2021, CAHAI completed its task, producing an outcome document that identified the possible elements of such a legal framework. According to the document, a legal instrument containing fundamental principles of protection of human dignity and the respect for human rights, democracy, and the rule of law is necessary for the development, design and application of AI systems. The instrument should provide for the establishment of a methodology for risk classification of AI systems, featuring the categories ‘low risk’, ‘high risk’ and ‘unacceptable risk’. AI applications presenting ‘unacceptable’ risks should be banned. In order to avoid unjustified bias, a provision on respect for equal treatment and non-discrimination should be included. Legal safeguards should guarantee at least the right to an effective remedy before a national authority, the right to be informed about the application of an AI system in a decision-making process, the right to choose interaction with a human, and the right to know that one is interacting with an AI system. However, specific issues like the manipulation of content (‘deep fakes’) should be dealt with in other sectoral instruments. A requirement to set up compliance mechanisms and national supervisory authorities should be considered. A non-binding impact assessment model could complement the legal instrument.
A successor to CAHAI – the Committee on Artificial Intelligence (CAI) – was set up for the period 2022-2024. According to its terms of reference, the CAI is instructed to set up an international negotiation process to develop a legal framework on the development, design and application of AI. The framework should be based on Council of Europe standards of human rights, democracy and rule of law, and should be conducive to innovation. The framework is to be drafted by 15 November 2023 and finalised by the time the CAI is wound up in 2024. In fulfilment of its terms of reference, the CAI should coordinate its work with other intergovernmental committees and Council of Europe entities, base its work on strong evidence and an inclusive consultation process, including with international and supranational partners, and take account of the CAHAI outcome document.
On 4-6 April 2022, the CAI held its inaugural meeting where a chair, vice-Chair and bureau were elected. On 30 June 2022, the Council of Europe Committee of Ministers instructed the CAI ‘to proceed speedily with the elaboration of a legally binding instrument of a transversal nature (‘convention’/ ‘framework convention’) on artificial intelligence based on the Council of Europe’s standards on human rights, democracy and the rule of law, in line with its terms of reference, focused on general common principles, conducive to innovation, and open to participation by non-member States, while taking into account other relevant existing international legal frameworks or those under development’.
Subsequently, the Chair of the CAI circulated a 'zero draft' of the future (framework) convention which lays down fundamental principles and rules aimed at ensuring that the design, development and application of AI systems are consistent with respect for human rights, the functioning of democracy and the observance of the rule of law. The future (framework) convention will have to be implemented in the respective jurisdictions of the Parties. The scope covers both public and private providers and users of AI systems, but excludes AI systems relating to national defence. The zero draft proposes to include the following provisions:
·purpose and scope of the (framework) convention;
·definitions for an AI system, lifecycle, provider, user and ‘AI subject’;
·certain fundamental principles, including procedural safeguards and rights for AI subjects that would apply to all AI systems, irrespective of their level of risk;
·additional measures for the public sector as well as AI systems posing ‘unacceptable’ and ‘significant’ levels of risk identified on the basis of a risk and impact assessment methodology (to be set out later in an annex to the convention);
·follow-up and cooperation mechanism between the parties;
·final provisions, including a possibility for EU Member States to apply EU law in their mutual relations for matters covered by the convention and a possibility for the Union to accede to the convention.
The zero draft will be discussed at CAI plenary sessions in Strasbourg planned for 21‑23 September and 23-25 November 2022. Four meetings are also scheduled in 2023 and one meeting in 2024.
•Reasons for and objectives of the proposal
This recommendation for a decision is submitted to the Council pursuant to Article 218 of the Treaty on the Functioning of the European Union (TFEU). The decision would authorise the opening of negotiations on behalf of the Union for a future Council of Europe convention on AI, human rights, democracy and the rule of law, to adopt negotiating directives and to nominate the Commission as the Union negotiator.
Article 3(2) TFEU provides that the Union has exclusive competence ‘for the conclusion of an international agreement… in so far as its conclusion may affect common rules or alter their scope.’ An international agreement may affect common rules or alter their scope where the area covered by the agreement overlaps with or is covered to a large extent by Union law.
In this regard, EU law provides for a comprehensive set of common single market rules for products and services for which AI systems can be used. Furthermore, the development and use of certain AI systems may affect the exercise of fundamental rights enshrined in the EU Charter of Fundamental Rights and secondary EU legislation, including, for example, the right to physical and mental integrity, protection of personal data, privacy, non-discrimination, freedom of expression and information, the presumption of innocence, fair trial and effective remedy. Such development and use may also affect the values on which the Union is based, including human dignity, democracy and the rule of law.
Moreover, to assess whether an area is largely covered by EU law, account must be taken not only of EU law as it currently stands in the area concerned, but also of its future development, in so far as that is foreseeable at the time of that analysis. The area covered by the future Council of Europe convention on AI is of direct relevance to such foreseeable future developments, taking into account the Commission proposal for a regulation on AI (‘the AI act’), presented in April 2021 and currently being negotiated by the European Parliament and the Council.
Negotiations for the future Council of Europe (framework) convention on AI do relate to matters of Union competence and there is a very significant overlap between the zero draft convention and of the proposed AI act in terms of their of the scope, nature and content. In view of this, it is important that the negotiations are conducted on behalf of the Union in such a way as to ensure that the consistency and uniformity of the EU rules for AI and the proper functioning of the system they establish are not undermined and that the future Council of Europe convention is fully consistent with existing and future EU law in the area.
•Consistency with existing policy provisions in the policy area
On 21 April 2021, the Commission proposed a regulation laying down harmonised rules on the placing on the market, putting into service and use of AI systems in the Union. The proposal has the twin objective of creating a single market to promote the development and uptake of AI, while addressing risks to safety, health and fundamental rights. The proposal is of a horizontal nature applicable to providers and users of AI systems, regardless of whether they are public or private entities. AI systems developed or used exclusively for military purposes are excluded from the scope.
The proposal lays down proportionate requirements and obligations limited to the minimum required to address risks to health, safety and fundamental rights, without hindering AI technological development or disproportionately increasing financial and administrative burdens on operators. For this purpose, it follows a risk-based approach classifying AI systems into different categories, namely ‘unacceptable’, ‘high’, ‘transparency-related’ and ‘low’ or ‘minimal’ risk.
·A ban is proposed for AI systems considered to be incompatible with EU values and a clear threat to the safety, livelihood or rights of people, posing an ‘unacceptable risk’ (e.g. social scoring by public authorities, harmful manipulative AI practices, real-time remote biometric identification for law-enforcement purposes in publicly accessible spaces, subject to some limited exceptions).
·‘High-risk’ AI systems posing significant risks to health, safety or fundamental rights must undergo conformity checks and comply with certain mandatory requirements (e.g. in relation to risk management, data quality, documentation, transparency, human oversight, accuracy, cybersecurity) before those systems can be placed on the market or used in the Union. Proportionate and effective monitoring and oversight obligations are also placed on providers and users, clearly defining their roles and responsibilities across the value chain without affecting other obligations under sectoral legislation.
·For AI systems posing a ‘transparency-related risk’, people should be informed when they are interacting with or being exposed to such systems (e.g. chatbots, emotion recognition and biometric categorisation systems, deep fakes).
·All other systems that pose ‘minimal’ or ‘low risk’ to health, safety and fundamental rights are permitted without further restrictions, but providers can choose to comply with voluntary codes of conduct.
The proposal is being discussed by the Council and the European Parliament in accordance with the ordinary legislative procedure.
The negotiations on the future Council of Europe convention on AI, human rights, democracy and the rule of law should ensure that the agreed provisions are compatible with EU law and Member States’ obligations under it, taking account of its future development and the proposed AI act. It will also be necessary to ensure that the Council of Europe convention includes a disconnection clause allowing the EU Member States that become Parties to the convention to regulate relations among themselves on the basis of EU law. The proposed AI act, as it evolves in the legislative procedure and negotiations by the co-legislators and eventually in its final (adopted) form, should be preserved amongst the EU Member States. The European Union should also be able to accede to the convention as a party.
•Consistency with other Union policies
The recommendation is relevant for other ongoing or planned Commission initiatives that aim to address problems posed by the development and use of AI systems, identified in the White Paper on AI. These other initiatives include revision of sectoral product legislation (e.g. the Machinery Directive, the General Product Safety Directive) and upcoming initiatives that address liability issues related to new technologies, including AI systems.
The recommendation is also consistent with the Commission’s overall digital strategy aimed at promoting technology that works for people, one of the three key objectives in the policy vision set out in the Communication ‘Shaping Europe’s digital future’. It aims to ensure AI is developed in ways that respect human rights and earn people’s trust, making Europe fit for the digital age and turning the next 10 years into the Digital Decade'.
The proposed recommendation also significantly strengthens the Union’s role in shaping global norms and standards and in promoting trustworthy AI that is consistent with the Union's values and interests. It provides the Union with a powerful basis for strengthening EU digital diplomacy and for negotiating, with other European and non-European countries, the first legally binding international agreement on AI, based on shared values and principles.
2.LEGAL BASIS, SUBSIDIARITY AND PROPORTIONALITY
Article 218(3) of the Treaty on the Functioning of the European Union (TFEU) provides that the Commission shall submit recommendations to the Council, which shall adopt a decision authorising the opening of negotiations and nominate the Union negotiator. Under Article 218(4) of the TFEU, the Council may address directives to the negotiator.
•Subsidiarity (for non-exclusive competence)
The proposed AI act lays down harmonised rules for AI systems in the Union, thereby preventing Member States from imposing further restrictions, unless explicitly authorised by the Act. Under Article 3(2) of the TFEU, the Union has exclusive competence for the conclusion of international agreements in so far as they may affect common rules or alter their scope. It follows, therefore, that for all matters relating to the design, development and application of AI systems, covered to a large extent in EU law or to be covered following the foreseeable future development of EU law and in particular the proposed AI act, the Union should have a competence to negotiate the future Council of Europe convention on AI.
This initiative does not go beyond what is necessary to achieve the policy objectives at stake.
•Choice of the instrument
The choice of the instrument is a recommendation for a Council decision, as stipulated in Article 218(3) and (4) TFEU.
3.RESULTS OF EX POST EVALUATIONS, STAKEHOLDER CONSULTATIONS AND IMPACT ASSESSMENTS
•Ex post evaluations/fitness checks of existing legislation
•Collection and use of expertise
The Commission has taken account of views expressed by Member State experts in discussions in the relevant Council working group to prepare for the negotiations.
•Regulatory fitness and simplification
The proposed AI act and the Council of Europe convention under consideration both seek to minimise risks and ensure a high level of protection of fundamental rights that might in certain circumstances be negatively affected by the development and use of AI, while also recognising the potential of AI to protect and facilitate the exercise of those rights in the digital environment and to improve societal and environmental wellbeing and technological progress.
There are no budgetary implications envisaged for the Union budget.
•Implementation plans and monitoring, evaluation and reporting arrangements
The negotiation process is expected to last until 2024, after which convention may be signed and concluded.
Recommendation for a
authorising the opening of negotiations on behalf of the European Union for a Council of Europe convention on artificial intelligence, human rights, democracy and the rule of law
THE COUNCIL OF THE EUROPEAN UNION,
Having regard to the Treaty on the Functioning of the European Union, and in particular Article 218(3) and (4) thereof,
Having regard to the recommendation from the European Commission,
(1)In 2021, the Council of Europe Committee of Ministers set up a Committee on Artificial Intelligence (CAI) for the period 2022-2024, tasked with setting up an international negotiation process to develop a legal framework on the development, design and application of artificial intelligence (AI), based on the Council of Europe’s standards on human rights, democracy and the rule of law, and conducive to innovation.
(2)On 30 June 2022, the Council of Europe Committee of Ministers instructed the Committee on Artificial Intelligence to proceed speedily with the elaboration of a legally binding instrument of a transversal nature (‘convention’/ ‘framework convention’) on artificial intelligence based on the Council of Europe’s standards on human rights, democracy and the rule of law, in line with its terms of reference, focused on general common principles, conducive to innovation, and open to participation by non-member States, while taking into account other relevant existing international legal frameworks or those under development.
(3)Subsequently, the Chair of the Committee on Artificial Intelligence proposed a zero draft of the (framework) convention that would apply to the design, development and application of AI systems. The zero draft includes: provisions on purpose and object, scope, definitions, fundamental principles including procedural safeguards and rights applicable to all AI systems irrespective of their level of risk, additional measures for AI systems in the public sector and for AI systems posing ‘unacceptable’ and ‘significant’ levels of risk, a follow-up and cooperation mechanism; final provisions, including a possibility for the Union to accede to the convention; and an appendix, under development, on a methodology for risk and impact assessment of AI systems.
(4)The Union has adopted common rules that will be affected by the elements considered for the Council of Europe convention. This includes in particular a comprehensive set of rules in the area of the single market for products and services for which AI systems can be used, as well as rules on the protection of fundamental rights enshrined in the EU Charter of Fundamental Rights and implemented in secondary Union legislation, considering that those rights are likely to be adversely affected in certain circumstances by the development and use of certain AI systems.
(5)Moreover, on 21 April 2021, the Commission submitted a legislative proposal for a regulation laying down harmonised rules for AI, currently being negotiated by the European Parliament and the Council. The envisaged Council of Europe convention to a large extent overlaps with the legislative proposal in its scope, since both instruments aim to lay down rules applicable to the design, development and application of AI systems, provided and used by either public or private entities.
(6)Therefore, the conclusion of the Council of Europe convention under consideration may affect existing and foreseeable future common Union rules or alter their scope within the meaning of Article 3(2) of the Treaty on the Functioning of the European Union.
(7)In order to protect the integrity of Union law and to ensure that the rules of international law and Union law remain consistent, it is necessary that the Commission be authorised to negotiate on behalf of the Union the Council of Europe convention on artificial intelligence, human rights, democracy and the rule of law.
(8)The European Data Protection Supervisor was consulted in accordance with Article 42(1) of Regulation (EU) 2018/1725 of the European Parliament and of the Council and delivered an opinion on …
HAS ADOPTED THIS DECISION:
The Commission is hereby authorised to negotiate, on behalf of the Union, the Council of Europe (framework) convention on artificial intelligence, human rights, democracy and the rule of law.
The negotiating directives are set out in the Annex.
The negotiations shall be conducted in consultation with [name of the special committee
to be inserted by the Council].
This Decision is addressed to the Commission.
Done at Brussels,
For the Council