Opinion of the European Committee of the Regions – White Paper on Artificial Intelligence — A European approach to excellence and trust

(2020/C 440/14)

Rapporteur: Guido RINK (NL/PES), Member of the Executive Council of Emmen Reference document: White Paper on Artificial Intelligence — A European approach to excellence and trust COM(2020) 65 final

POLICY RECOMMENDATIONS

THE EUROPEAN COMMITTEE OF THE REGIONS

General comments

1.

notes that on 19 February 2020, the European Commission (the Commission) published its digital strategy for 2020-2025. With its White Paper on ‘Artificial Intelligence’ (AI), the Commission sets out its vision on key policies and investment in the field of AI. The Commission believes that Europe can become a global leader in AI;

2.

reminds that the AI White Paper is linked to the Communications on Europe’s Digital Future of Europe (1) and the European Data Strategy (2);

3.

points out that the Commission has declared that its objective is to define an approach for developing and using AI in a way that is human-centred, ethical, inclusive and secure, and respects the values with which the EU wishes to associate itself;

4.

recalls that the AI White Paper has two main components: promoting the development and uptake of AI and regulating ethical aspects and trustworthiness;

5.

welcomes the White Paper on Artificial Intelligence — A European approach to excellence and trust. It recognises the importance of the discussions on AI ahead of Commission policy-making in this area. The CoR does point out that it would have liked the role of regional and local authorities as a key partner, producer, user and promoter of innovation in the development of AI also to have been stressed in the White Paper;

6.

recalls, in this context, the CoR opinion on Artificial Intelligence of 6-7 February 2019 (3);

7.

points out that AI is already part of our lives and will play an increasingly important role in transforming our societies. It offers great potential for European society, businesses and the public. Innovation in the field of AI does not just make a contribution to the economy; it also helps to solve societal and environmental challenges;

8.

considers that the application of AI is important not only at national level but also, and above all, at a local and regional level. Local and regional authorities have a role to play in, among other things, stimulating AI ecosystems in their own regions and in promoting investment — and themselves investing — in AI;

9.

underlines that AI is a complex policy area that does not stand alone. It touches upon a number of other policies pursued by the Commission;

10.

points out that the AI White Paper and its policies should be linked to the Communication on Shaping Europe’s digital future presented in parallel with the White Paper on AI and the Communication on a European Data Strategy. In addition, the White Paper should be linked to a number of other Commission policy areas, such as the Skills Agenda for Europe (4) and the Recommendation on a common Union toolbox for the use of technology and data to combat and exit from the COVID-19 crisis, in particular concerning mobile applications and the use of anonymised mobility data (5);

11.

stresses that the CoR’s recommendations should, where relevant, be considered in conjunction with these European policy-making documents;

12.

welcomes the Commission’s approach, which has at its heart the development of human-centric AI and the corresponding goal to take the lead in setting ethical standards. In this connection, the Commission rightly points out that the benefits of AI depend on public confidence in AI;

13.

insists, however, that the debate on future policy measures to build up trust in AI should also address questions about the ownership of data, algorithms and platforms, about steps to ensure public values on platforms and about who ultimately gains most from AI applications and who pays for them (and whether it is socially acceptable);

14.

points out that, bearing in mind that AI affects a large number of other policy areas, there is a risk that consistency between these policy areas may be diluted and that those policy areas develop in a vacuum;

15.

suggests that the Commission define a roadmap and approach to promote consistency between different policy fields;

Making use of opportunities

16.

emphasises that European policies should strongly promote and invest in (steps to organise and stimulate) multidisciplinary local and regional networks of people, administrations, knowledge institutions and businesses. The CoR believes that the innovation hubs proposed by the Commission would play a central role here;

17.

points out that AI has the potential to help solve societal challenges in areas such as health (with the fight against the global COVID-19 crisis as the main recent example), security, climate, mobility and transport, social support, high-tech industries, retail, agriculture, tourism and public services;

18.

draws attention to the fact that AI has the potential to create new jobs and generate new entrepreneurship. The Commission rightly acknowledges that an important precondition is that the public should be able to trust technology. A strategic EU framework based on fundamental values should boost confidence and encourage firms to develop AI solutions;

19.

agrees that a level playing field in the European market is important. Particular attention should be paid to the AI access for small and medium-sized enterprises (SMEs) and start-ups. They are key engines for local and regional economies. European Digital Innovation Hubs and future regulatory frameworks and policies have a decisive role to play in promoting equal opportunities of AI access, especially through supporting micro, small and medium-sized enterprises (MSMEs) in accessing AI;

20.

understands that the Commission is committed to making use of Europe’s strong position in (existing) industrial and professional markets (6). It wishes to emphasise that this strategy should not however come at the expense of productivity, innovation and the DNA of local and regional ecosystems;

21.

stresses that local and regional authorities work on a wide range of social tasks in these ecosystems. They can play an important role in accelerating the impact of innovation in these tasks;

22.

underlines that local and regional authorities are best placed to help create an environment propitious to boosting investment in AI at the various points in the value chain in the coming years and fostering trust in AI. This is because they are closest to AI use in practice and can drive local multidisciplinary networks;

23.

considers that significant EU support will be needed to stimulate private and public investment and provision will have to be made for resources from the Digital Europe Programme, Horizon Europe and the European Structural and Investment Funds to meet the needs of local ecosystems;

24.

is of the view that subsidies should be channelled to specific AI applications that are suitable for scale-up. In order to maximise the opportunities offered by AI, small-scale projects should be promoted and the financial burden of these projects limited, making it more attractive for local and regional authorities to participate in EU-funded projects;

25.

supports the approach adopted for the development of AI within digital innovation hubs. Here, subsidies for digital innovation hubs should focus on fostering local and regional ecosystems;

26.

is pleased that the Commission is committed to the development of products based on AI in the public sector (7);

27.

considers, however, that this commitment to the development of AI in the public sector should not be limited to AI technology that has already been proven in pre-defined sectors. Indeed, local and regional authorities can play an important role in speeding up (new) AI technology as ‘launch customers’;

28.

requests the Commission to call on all public authorities, including the local and regional ones, to undertake strict fundamental rights impact assessments for AI systems deployed in the public sector. Authorities should avoid having recourse to any AI surveillance technologies, especially in times of urgencies, before the results of the impact assessment are known and necessary solutions adopted;

29.

calls for the development of a procurement framework and legal tools, such as standard procurement conditions, that offer local and regional authorities possibilities for action and allow them to make optimal use of opportunities;

30.

underlines that the value of data is in the re-use thereof, including in AI applications. Cross-sectoral, secure data sharing fosters AI innovation. The CoR urges the Commission to use the General Data Protection Regulation (EU) 2016/679 and to assess the pertinence of a further, AI-specific regulatory framework, for the easy and reliable sharing and re-use of data between businesses and public authorities (B2G), especially when it can serve the public interest, such as in the case of the COVID-19 pandemic. This has also shown how important it is for public authorities to use business data. The potential should be exploited as it can reduce the administrative burden on businesses and in many cases, in the context of the platform economy for example, it is the only way in which public authorities can meet their legal obligations;

31.

considers it necessary to invest at European level in shared, safe, basic infrastructure, data architecture and quality standards as regards safety, data sets and statistics. This lowers the entry threshold for the use of AI applications and increases digital trust, boosting the development and use of AI;

32.

firmly believes that AI applications developed with public funds should also be given back to society as much as is possible. Such AI applications should in principle be developed for instance under open source conditions, while also recognising the need for funding for development;

Legislation and policy: an approach involving a learning process

33.

notes with interest the progress made in the definition of AI, as this is reflected in the updated definition (8) provided by the Independent High-Level Expert Group (AI HLEG), set up by the European Commission. It believes that this definition better reflects the technical capabilities of AI. However, the definition of AI should be an ongoing process. It should take into account the context in which AI operates and it should keep pace with societal developments in this field and not lose sight of the link between the Commission’s ecosystem of excellence and the ecosystem of trust;

34.

considers that an AI system consists of a combination of technical elements that link data, algorithms and processing power to social practices, society, identity and culture. The definition of such a dynamic socio-technical aggregate should therefore be regularly updated to accurately reflect AI’s ever-growing societal impact, while identifying fast changing AI-related challenges and opportunities;

35.

notes that, according to the Commission, the key elements of the ecosystem of trust are the protection of fundamental rights and consumers’ rights, transparency in using the aforementioned technical elements and rules on liability and product safety. These core elements form the basis for a future EU regulatory framework for AI;

36.

points out how machine learning in the AI industry is based on human programming, which might lead to bias at a large scale. Therefore, calls for the EU to implement mechanisms to ensure gender, race and worldview equality and inclusivity in the development and applications of AI technology;

37.

draws attention to the fact that criminal law offers an important way of setting standards to protect against serious violations of people’s fundamental rights and unlawful and clandestine surveillance of the public through AI applications;

38.

agrees with the Commission that AI applications are already regulated by a broad, high-quality framework of EU legislation and ethical principles. Key examples are legislation on product safety and liability, consumer rights and fundamental rights, anti-discrimination legislation and the right to personal data protection;

39.

emphasises the need to recognise that the existing legislative framework is not specifically tailored to AI applications, which means that there is currently a ‘gap’ in the law. There should be careful examination of which areas need additional regulation to boost peoples’ trust in AI. A pivotal element of any future regulatory framework applying to AI is the introduction of safeguards to ensure that AI is free from bias and does not reproduce discrimination on grounds of gender, ethnicity age, disability or sexual orientation;

40.

also points to the need that none of the official EU languages is discriminated and made vulnerable by the use of AI, and that there are data and language sets available in all EU languages;

41.

further insists on the importance of gender-sensitive coding and calls for the equal participation of all genders in the design, implementation, evaluation and debate on ethics and norms of AI-related technologies. Boosting girls’ and women’s participation in the field of science, technology, engineering, arts and mathematics (STEAM) is essential for their full inclusion in AI-related processes in particular and the digital economy in general;

42.

underlines that this regulatory framework should ensure enough of the above-mentioned key elements, while allowing room and flexibility for innovation. In so doing, we should be aware of the challenges posed by the interpretation and operation of AI systems and the outcomes and social impact caused by such systems;

43.

notes that AI is a technology that does not stand alone but is linked to other technologies and sciences, such as behavioural science, quantum computing, the internet of Things, deployment of 5G and 6G networks, business cases and digital platforms;

44.

considers it necessary to take into account the fact that AI is a developing technology that has not yet been developed to the full and is not fully embedded in society;

45.

points out that politicians, policy-makers and society face a fundamental challenge here: How do we ensure that the relationship between desired and unwanted outcomes and impact goes the right direction? And how do we ensure that there is sufficient room to exploit the opportunities of AI and increase people’s trust in it? With the advancement of AI applications, especially predictive algorithms, in many public and government institutions (judicial bodies, law enforcement, military etc.), there is an urgent need for a regulatory framework foreseeing the strict test of necessity and proportionality, providing appropriate safeguards and remedies and clearly defining the responsibilities and accountability as well as proper public oversight;

46.

stresses that regional and local authorities have an important role to play in shaping AI legislation and policies. They are closer to the public and therefore have more useful data than national governments. Regional and local authorities should therefore be more involved in the development of policies and legislation flowing from the White Paper;

47.

supports the Commission’s view that, in view of how fast AI is developing, the regulatory framework should leave room to cater for further developments. This calls for adaptive legislation and an adaptive legislative process. It also calls for a critical approach on the part of the Commission to the operation of its own system and to social innovation in this system;

48.

insists in this context that regional and local authorities should have sufficient scope to experiment (including policy prototyping) and to learn how policies for these fundamental challenges can be shaped most efficiently and effectively;

49.

urges the Commission to develop a procedural policy framework aimed at using to the fullest extent already existing rules, but also at further developing future AI policy and the EU regulatory framework over time. Such a policy framework should also coordinate efforts at European, national and regional level and ensure cooperation between the private and public sectors. To this effect, the creation of an EU Agency on Artificial Intelligence would contribute to the effective oversight and coordination of AI-related issues amongst all levels of governance, from European to local;

50.

supports the Commission’s approach of devising additional regulation for high-risk AI applications;

51.

considers, however, that the main criteria for determining whether AI is ‘high-risk’ are: the extent to which humans can influence decision-making and the impact of those decisions on people’s rights and actual behaviour;

52.

calls for future legislation and policy to focus in particular on the transparency and ‘explainability’ of algorithms and the accountability, (9) fairness and responsibility of those deploying AI, in particular where people’s rights and actual behaviour are affected or influenced;

53.

points out that people have the right to know, in plain language, on the basis of which data and algorithms their rights and actual behaviour are affected, so that they can defend themselves fairly and, if necessary, obtain effective legal protection. In addition, the decision-making system should include the option of having an issue dealt with by a person. Authorities’ role should be emphasised where appropriate;

54.

considers the use of AI technologies highly relevant also when it comes to working conditions and workers’ well-being. It joins to this effect the call of European Social Partners for ‘data minimisation and transparency along with clear rules on the processing of personal data to limit the risk of intrusive monitoring and misuse of personal data’ (10), in order to ensure the respect of human dignity. In this context, the Committee agrees with the importance of enabling workers ’ representatives to address issues related to data, consent, privacy protection and surveillance, linking the collection of data to a concrete and transparent purpose and ensuring transparency when AI systems are used in human-resource procedures;

55.

agrees with the training data requirements that the Commission is envisaging. Keeping records is one way of verifying compliance with rules. However, the CoR considers that the administrative burden entailed should be as limited as possible;

56.

supports the Commission’s view that ‘the objective of trustworthy, ethical and human-centric AI can only be achieved by ensuring an appropriate involvement by human beings in relation to high-risk AI applications’;

57.

insists that future legislation and compliance monitoring should apply to the entire life cycle of the AI application;

58.

supports the Commission’s view that human oversight helps to ensure that ‘human autonomy’ is preserved, however calls on the European Commission to assess ethical risks, in particular related to biases while using AI and propose clear solutions;

59.

calls for AI regulation to consider the short-term and long-term environmental impact of the usage of these technologies throughout their lifecycle and across the entire supply chain;

60.

has taken note of the draft report by the European Parliament’s Committee on Legal Affairs (11). In particular, the CoR notes the idea that each Member State should establish a national supervisory authority responsible for ensuring, assessing and monitoring compliance, and for enabling discussion and exchange of points of view in close cooperation with the concerned stakeholders and the civil society;

61.

supports the Commission’s view that an objective, prior conformity assessment would be necessary to verify and ensure that mandatory requirements applicable to high-risk applications are complied with;

62.

agrees with the Commission that the consequences of using remote biometric identification AI systems can vary considerably depending on the purpose, context and scope of the use;

63.

requests that a binding, unambiguous quality framework be developed to guide such major AI applications. Such a framework should focus on standards and practices that prevent unlawful discrimination and stigmatisation of individuals and populations. The CoR supports the Commission’s approach of carrying out a wide-ranging debate on this issue;

64.

believes that ethics, alongside legislation, have an important role to play in the design of AI (ethics-by-design). We must be aware that ethics are not linked to a specific technical tool (such as AI), but are the social context and regulatory framework within which the technical tool is applied;

65.

considers that the future policy for the ecosystem of trust will benefit from a broader approach than simply drafting legislation for high-risk AI. Future policy will require a continuous, systematic socio-technical approach, looking at technology from all perspectives and through different lenses. This calls for a multidisciplinary approach to policy-making and regulation, where policy-makers, academics from different fields, social partners, businesses and local and regional authorities permanently cooperate and monitor developments and share their findings with the same level of transparency;

Knowledge development

66.

points out that, to promote the transition to a society in which AI plays an important role, it is necessary for the forthcoming proposals for an Updated Skills Agenda and for a Digital Education Action Plan to extend to the entire line of learning, from primary education, secondary education and secondary vocational education to academic education and life-long learning. Digital education curricula should promote active citizenship, sharpen critical thinking and empower people from an early stage to handle growing interaction with AI;

67.

calls for educational and training measures to also focus on boosting the digital skills of the public and professionals, both in the education system and in ongoing vocational training. With the digital revolution, the number of people employed in digitally relevant professions is expected to increase significantly. At the same time, life-long technological learning in AI is essential not only for the technical professions based on STE(A)M studies, but for all workers (including public administration employees) who will need AI knowledge in many other areas of activity. Thus, training should focus not only on current labour market needs linked to AI, but also on technological literacy for all workers, which allows for the adaptation to a long-term approach to training needs in the field of AI;

68.

underlines that training politicians and policy-makers, not only in the use of AI, but also in ethical provisions and standards is crucial and will improve democratic decision-making. The CoR recommends trainings that entail updates at a relatively high level and has two objectives: (1) to be able to communicate with the market on a relatively equal footing; and (2) to manage the impact of AI on society and the democratic process;

69.

stresses that, in order to ensure that AI reflects our fundamental values and rights and that gender-biased programming is avoided, it is crucial to diversify the technology sector and encourage students, in particular girls, to enrol in STE(A)M courses;

Multilevel governance and public-private cooperation

70.

agrees there is a need for a common European approach to AI in order to achieve sufficient scale and avoid fragmentation of the single market;

71.

stresses, however, that a European approach to governance must be based on an open, inclusive and decentralised society, in which everyone has the possibility to be involved, creative and entrepreneurial;

72.

argues that the main question here is how politics, the public and SMEs can be effectively involved in developing AI applications, ethics and regulation, and in what way understanding the expected return from the ecosystems can play a role in the (re)calibration of policies and regulations;

73.

highlights the important role of decentralised networks of people and local economies in involving politics, the public and businesses in the development of AI applications, ethics and regulation. Indeed, the strength of local communities and networks lies in open, interconnected and adapted local and regional cooperation fostering innovation and the development of new economies;

74.

believes that the future European policy framework should coordinate efforts at European, national and regional level, promote knowledge sharing and ensure cooperation between the private and public sectors. This calls for multi-level governance, linking up local, regional, national and European networks.

---

(1)  COM(2020)67.

(2)  COM(2020)66.

(3)  SEDEC VI/046 (OJ C 168, 16.5.2019, p. 11).

(4)  COM(2016) 381.

(5)  Recommendation (EU) 2020/518.

(6)  Chapter 2 of the AI White Paper.

(7)  Chapter 4F.

(8)  https://ec.europa.eu/newsroom/dae/document.cfm?doc_id=56341

(9)  Maranke Wieringa, What to account for when accounting for algorithms, Utrecht University, 20 January 2020.

(10)  European Social Partners Autonomous Framework Agreement on Digitalisation (https://www.etuc.org/system/files/document/file2020-06/Final%2022%2006%2020_Agreement%20on%20Digitalisation%202020.pdf).

(11)  Draft report by rapporteur Ibán García del Blanco on 21 April 2020 with recommendations to the Commission on a Framework of ethical aspects of artificial intelligence, robotics and related technologies (2020/2012 (INL)).