AI services and applications are increasingly influencing daily life and business operations, and have become increasingly important for the future viability of many companies worldwide. As early as 2021, the EU Commission published the first draft of an AI Act intended to strengthen trust in AI technologies. However, after new services entered the market that included new chat and image AI technologies, the Parliament felt compelled, due to discussions, to also include such generative AI in the regulation, leading to delays in the legislative process. Now, on April 17, 2023, the European Parliament announced that the Committee on the Internal Market and Consumer Protection (IMCO) and the Committee on Civil Liberties, Justice and Home Affairs (LIBE) will vote on April 26, 2023, on the draft report for the proposal of an AI Act (Artificial Intelligence Act).

The most important points in brief

• The EU Commission wants to regulate the placement on the market, development, and use of AI in a uniform law and strengthen trust in AI systems.
• The EU's AI Act would be the world's first law that could influence the development and deployment of AI.
• The planned regulation pursues a risk-based approach, so that higher or lower requirements apply depending on the risk, for example, with regard to safety or transparency obligations.
• Once the European Parliament has finalized its position on the draft regulation, it will enter into negotiations with the Council of the EU.

EU Commission legislative proposals for regulating AI

Artificial intelligence is a technological function that enables computers to perform human-like intellectual tasks. Recognizing the potential of AI systems many years ago, the Commission began promoting research in this field early on. In this context, on April 21, 2021, it published a draft of the Artificial Intelligence Act. This is a legislative proposal for the uniform regulation of AI within the EU. With this proposal, the Commission aims to ensure a high level of protection and security for individuals and businesses in the EU by restricting and regulating the use of AI systems in specific types and applications.

To this end, the legislator has divided AI systems into four risk categories in the draft regulation: prohibited systems, high-risk systems, risk-based systems, and non-risk systems. For example, high-risk systems, such as those used in autonomous vehicles, medical diagnostic systems, or for monitoring public parking spaces, are subject to stricter regulation and must meet certain requirements before they can be used in the EU. In contrast, the regulation provides for less stringent requirements for non-risk systems, such as chatbots or voice assistants.

For all AI systems, it is paramount that they are transparent in order to strengthen consumer trust in the new technology. Furthermore, the law sets specific requirements for data security, human oversight, verification and certification of AI systems, and the reporting of accidents and incidents related to AI systems. In addition, the law stipulates that liability for any damages that arise in connection with the use of AI systems shall generally lie with the respective AI developer and not with the user.

Final vote by the committees of the EU Parliament

Before the law can enter into force, it must first be adopted by the member states and the EU Parliament. The draft is currently in the EU Parliament. The parliamentary committees have worked on a draft report containing a revised proposal for an AI regulation. After a joint vote by the committees involved, a vote in Parliament will follow.

As a result of the emergence of new chat and image AI services, legislative proceedings were delayed due to a series of amendments from the EU Commission.

Parliament most recently informed that the current draft report by the IMCO and LIBE rapporteurs contains significant changes to the AI Act following thorough negotiations, such as a ban on predictive policing, additions to the list of standalone AI classified as high-risk, as well as a strong and comprehensive role for the planned AI Office.

Furthermore, a stronger alignment with the General Data Protection Regulation (GDPR), as well as greater involvement of stakeholders in various areas and the introduction of specific provisions concerning general artificial intelligence are proposed.

The permissibility of real-time biometric identification in public spaces was also discussed. However, both the Parliament and the Member States agreed early on that this should be prohibited in principle. Recently, however, calls for exceptions have been heard not only in the Council of the European Union.

Similar to the GDPR, the AI Regulation provides for penalties, particularly fines, in the event of violations. Specifically, it provides for fines of up to 30 million euros or up to 6% of a company’s global annual turnover.