The European Union Proposed Artificial Intelligence Act: Prospects And Challenges

By Halima Ummi Ismail

Introduction

Artificial Intelligence (AI) systems have crawled into the lives of humans in modern generation. The influence of AI in our daily activities is so obvious to the extent that these technologies have the ability to study humans through the data its generate thereby, predicting the exact personality of people and the nature of contents which they require to see. AI technologies have various applications in different part of life. These technologies proffer a lot of benefits to human society both socially and economically. AI technologies are now being adopted in divergent human endeavours and fields, from the public, health, environmental, finance, agricultural, home affairs, and legal systems have proved to be very useful in providing a wide range of services thereby, optimizing the abilities and operations of humans by making their works faster and much more efficient. AI systems are also known to have contributed to the development of the various economies.

Despite the numerous potentials and benefits which AI systems have to offer, they are not without loopholes. The use of AI technologies may in one way or the other affects humans negatively. AI systems may likely collide with the fundamental human rights of people including infringement on data privacy and security. These systems may likely be used in promoting discrimination, infringement on the dignity of humans and a host of other breaches. Also, the adoption of AI systems comes with serious risk factors. Hence, in a world where AI technologies are in the verge of dominating every sector of human activities, there is a need to create laws which seek to regulate AI systems and other related technologies.

In order to regulate AI technologies and ensure that people benefit from them, policy makers in some countries where the adoption of AI systems is high have made policies for regulation of the AI tools, however, most of these policies are not sufficient. As such, the European Commission in 2021 made a proposal for the creation of the Artificial Intelligence Act.[i] The proposed Act if passed into law will be the first of its kind.[ii] The European Union as a major regulator intends to enact the first regulatory framework, an initiative which is first of it kind globally,[iii] which will subsequently serve as the standard for regulatory framework of AI technologies worldwide just like the General Data Protection Regulation (GDPR).[iv] The Artificial Intelligence Act also aims at protecting the right of humans as well as the protection of personal data and other cyber affairs which may not necessarily be personal. The provisions of the proposed AI Act will have a significant effect in the deployment and development of AI technologies across the world thereby, giving the European Union the chance to unilaterally regulate the worldwide standards in relation to AI systems. This cannot be said to take effect forcefully, rather, the European Union’s consumer market which is very strong and its international dealings with other countries places it at the edge and gives it the power to influence the decisions of other countries.[v] However, the AI Act as of this moment is a proposal which may take a couple of years to conclude and bring it into existence as a law which governs AI technologies.

An Overview of the Proposed AI Act

Definition of AI

In ensuring that the definition of an AI system provide sufficiently clear criteria for distinguishing AI from more classical software systems, the compromise text narrows down the definition in Article 3(1) to systems developed through machine learning approaches and logic- and knowledge-based approaches.[vi]

Prohibited AI Practices

Article 5 of the proposed Act aims at prohibiting AI systems that are used in Exploiting vulnerable persons who are in this condition either due to their economic or social status. It also prohibits systems that can cause potential harm to such individuals, and AI systems that are used for social scoring both by public authorities and private individuals. As for biometric identification systems available at public places used by law enforcement agencies, such systems can only be allowed where they become strictly necessary by law for which these systems are exceptionally allowed for such purposes.

High Risk AI Systems

AI systems that are regarded as having high risks are those that belong to the category of AI that have been listed in the proposal as such which includes those that are used in determining the individual’s creditworthiness.[vii] They also include systems that are within the product safety regulations of EU which includes medical devices and toys, critical digital infrastructure, life and health insurance. The Act requires that the providers of such high Risk AI systems before putting them out in the market or before using them for service should register such systems in the wide database of European Union which is managed by the European Commission. Such providers are expected to abide by the provisions of the Act.

Limited Risk AI Systems

These includes emotion recognition systems, Chatbot systems that interact with humans, AI systems that generate or manipulate audios, images, or the content of videos, and biometric categorisation systems. These AI systems are subjected to very limited set of transparency obligations.[viii]

Minimal Risk AI Systems

These categories of AI systems are considered to have very little risk and can be used with very minimum obligations.  Example of such includes video games, spam filters, and so on. Since these systems do pose little or no harm at all to the safety and rights of citizens, the proposed Act permits the use of such applications with no restrictions and no requirements.

General Purpose AI Systems

The proposed Act has made provisions for AI systems that can be used for various purposes. These AI systems can also be classified into high-risk AI systems or its environments. This shows that some specific requirements that are used for high-risk AI systems will also be made applicable to general purpose AI systems. However, rather than applying these requirements directly, an act which is made for implementation will specify the application of such requirements in relation to general purpose AI systems based on consultation and detailed impact assessment and taking into account specific characteristics of these systems and related value chain, technical feasibility, market and technological developments.[ix]

Supervisory Authority

The proposed Act requires the appointment of National competent authorities for implementation of the act. It also made provisions for the designation of a National supervisory authority which will act as a market surveillance authority.  The European Artificial Intelligence Board is to ensure that the Act is consistently applied, and this should be chaired by the European Commission with the composition of the European Data Protection Supervisor, and representatives of the National Supervisory Authorities.[x]

Criticisms

The AI Act garnered the attention of all necessary stakeholders worldwide. This implies that the Act has succeeded in gaining both the support of some people as well as criticisms from many. Supporters of the Act are of the view that AI systems have so much influence on the lives of humans, and as such, should be regulated. AI systems controls the information we get online, it analyzes  our data and personalises the kind of output we get. It has so much control over every aspect of human endeavour, this paves way for the need to address the risks associated with these systems while at the same time, maintaining and promoting the development of AI systems. As such, in order to bring into harmony the promotion of AI systems and the need to protect mankind from the possible risks attached with these systems, the AI Act should be made applicable.

On the other hand, the critics of the proposed Act have opined that the technical and regulatory requirements set out by the Act are complex, rigid, and ambiguous in nature. Some critics from Civil Society Organizations have also criticised  the Act as having fall short of the protection of the fundamental human rights. These Organisations have resorted to calling out the attention of stakeholders on the need to make some changes to the proposed Act in order to suit the enforcement of fundamental human rights and to protect people from the likely harms and risks associated with certain AI systems. The Organisations were also of the opinion that the Act should place a total ban on social scoring systems, remote biometric identification in public places, all other biometric categorisation, emotion recognition systems, on all systems that can be used in predicting future criminal acts, and on all AI systems that may pose unacceptable risks to fundamental human rights.[xi]

One major concern of stakeholders in the tech industry with the provisions of the AI Act is in terms of definition which is a major area of concern for tech companies and businesses. They opined that the there is a need for the Act to make clarity on the definition of AI and the classifications of AI applications that are considered as high risk AI systems. To these companies, a narrower definition will be much better and easier to comprehend. TechUK, a body representing the UK tech industry, in its feedback submissions wrote that the current definition of AI by the Act is far ahead and beyond what will on a normal ground be tagged as intelligent.[xii] The obligations associated with the classification of AI into high risk AI according to major tech companies will serve as a barrier to start-ups and small-scale tech providers.

The AI Act: Way Forward

The European Union’s proposed AI Act is aimed at devising a regulatory framework for AI. The Act being a proposal is a work in progress which needs to pass through different proceedings both at the European Council and the European Parliament. These bodies need to scrutinize the Act and subject it to voting before it is passed into law. The AI Act has received mixed reactions from stakeholders and members of the public. Being a proposal, there is the likelihood of having some of its provisions altered and corrected to meet the necessary requirements.   As stakeholders continue to engage the Act through the drafting process, it is important that law and policy makers as well as the general public recognize the effects of the proposed Act, the responsibilities as well as the opportunities it seems to offer to the society. AI systems will continue to evolve and grow with time, as such, this brings the need to ensure that their activities with humans in human societies are duly regulated. Unlike the GDPR, the AI Act is expected to go through a rapid process of implementation.

Conclusion

AI systems are everywhere, they have become a part of human live and cannot be dealt away with. They influence the information we get online, the manner we carry out daily transactions, our work life, and so on. Some of these AI technologies have the potentials of putting humans at risk by harming its users. The rapid growth and disruptions of these technologies therefore, promotes the need to have a regulatory framework to put them in check, hence, the birth of the proposed AI Act. This Act is one of its kind being a novel idea of the European Union to set the regulatory standard for AI worldwide. The proposed Act however, is not without loopholes as various stakeholders have pinpointed their areas of concern. Despite the criticisms, it cannot be denied that the world needs an avenue to regulate AI technologies, and this is the major aim of the AI Act. As such, it is predicted that the Act will be thoroughly checked while taking notes of the necessary provisions that needs to be changed or altered, and at the end of the day, the European Union will present to the world the first AI Act as a binding law in member countries.

Halima Ummi Ismail is a penultimate law student of Bayero University, Kano. She is an avid researcher with a keen interest in technology law, and artificial intelligence. She can be contacted via email: ismailhalima99@gmail.com: LinkedIn: www.linkedin.com/in/halima-ummi-i-459267118

[i] Future of Life Institute ‘The AI Act’ https://artificialintelligenceact.eu/documents/ accessed January 19, 2023

[ii] Securiti ‘EU Artificial Intelligence Act’ https://securiti.ai/eu-artificial-intelligence-act/ accessed January 20, 2023

[iii] European Data Protection Supervisor ‘Artificial Intelligence Act: A welcome initiative, but ban on remote biometric identification in public space is necessary’ (2021) https://edps.europa.eu/press-publications/press-news/press-releases/2021/artificial-intelligence-act-welcomed-initiative_en accessed January 21, 2023

[iv] Uche Ewekukwa Ofodile ‘Businesses and the EU’s Proposed Artificial Intelligence Act’: Major Points of Controversy’ (2021) JURIST https://www.jurist.org/features/2021/12/17/businesses-and-the-eus-proposed-artificial-intelligence-act-major-points-of-controversy/ accessed January 21, 2023

[v] Alex Engler ‘The EU AI Act will have global impact, but a limited Brussels Effect’ (2022) Brookings https://www.google.com/amp/s/www.brookings.edu/research/the-eu-ai-act-will-have-global-impact-but-a-limited-brussels-effect/%3famp accessed January 19, 2023

[vi] Council of the EU ‘Artificial Intelligence Act: Council calls for promoting safe AI that respects fundamental rights’ (2022) https://www.consilium.europa.eu/en/press/press-releases/2022/12/06/artificial-intelligence-act-council-calls-for-promoting-safe-ai-that-respects-fundamental-rights/  accessed January 20, 2023

[vii] Frutz-Ulli Pieper ‘The AI Act- does this makes a turning point for regulation of Artificial Intelligence? An Overview’ Taylor Wessing https://www.taylorwessing.com/en/interface/2021/ai-act/the-ai-act–does-this-mark-a-turning-point-for-the-regulation-of-artificial-intelligence-an-overview accessed January 21, 2023

[viii] Ayush Patel ‘Dealing with the EU Artificial Intelligence Act’ (2022) https://towardsdatascience.com/dealing-with-the-eu-artificial-intelligence-act-40e7059c8210 accessed January 27, 2023

[ix] Sidley ‘Proposal for EU Artificial Intelligence Act passes next level- where do we stand and what’s next?’ (2022) https://www.sidley.com/en/insights/newsupdates/2022/12/proposal-for-eu-artificial-intelligence-act-passes-next-level accessed, January 21, 2023

[x] Shubhangi Agarwal ‘EU Artificial Intelligence Act-An Overview’ (202 TMT Law Practice https://www.legal500.com/developments/thought-leadership/eu-artificial-intelligence-act-an-overview/#_ftn16 accessed January 21, 2023

[xi] Natasha Lomas ‘Europe’s AI Act falls short on protecting fundamental rights, civil society groups Warns’ (2021) Techcrunch.com https://www.google.com/amp/s/techcrunch.com/2021/11/30/eu-ai-act-civil-society-recommendations/amp/ accessed January 22, 2023

[xii] Laurie Clarke ‘MEPs are preparing to debate Europe’s AI Act’ (2021) Tech Monitor https://techmonitor.ai/policy/meps-are-preparing-to-debate-europes-ai-act-these-are-the-most-contentious-issues  accessed January 27, 2023