In brief
A key step towards the adoption of the EU AI Act was reached on 2 February 2024 as the draft text received unanimous approval from the European Council’s main preparatory body. There are further votes to follow before the Act is adopted, but it’s looking likely that the final vote will take place in April and some substantive provisions of the Act could be in force soon after that, possibly by the end of the year. We set out below how the draft text has developed since the political agreement was announced in December 2023, and the remaining steps in the legislative process
What does the Act say, and what has changed in the latest draft?
The Act has been some time in the making, with hundreds of amendments tabled during the legislative process. There has been intense debate on many aspects of the Act and the balance it strikes between regulating AI and facilitating innovation, recently focused around whether and how the Act should regulate AI foundation models, the advanced generative AI models that are trained on large sets of data with the ability to learn and perform a variety of tasks, as well as over the use of AI in law enforcement.
In December 2023, the EU Parliament, Council and Commission announced a political agreement on the Act, and you can find a summary of the key provisions agreed in December 2023 in our post here EU reaches agreement on landmark AI regulation. EU legislators have since been working to reflect that political agreement in a final draft of the Act, ready for votes in the European Council and Parliament. A number of member states including Germany and France have been critical of the resulting agreement, but ultimately, all member states voted in favour at a key meeting of the European Council’s Committee of the Permanent Representatives of the Governments of the Member States to the European Union (COREPER).
The final compromise text has been published following the COREPER vote. The published text reflects the political agreement announced in December, but includes the detailed drafting on critical questions like the definition of an in-scope AI system. Some key developments include the following:
- A new provision has been added to the main Act, obliging the Commission to develop guidelines on the practical implementation of the Act, including:
- How the obligations on high-risk AI systems should be applied
- The AI practices prohibited under Article 5 of the Act
- The practical implementation of provisions relating to “substantial modification” of an AI system, and of the transparency obligations imposed on certain AI systems and on general purpose AI systems
- The relationship between the AI Act and other relevant EU law, including as regards consistency in their enforcement
- The application of the definition of an AI system
- The drafting of the recitals was not discussed in detail during the political negotiations. Recitals provide important guidance on key issues – including the definition of an AI system, what it means for an AI system to “materially influence” decision making, and how prohibited use cases should be interpreted – and have been updated in the published text.
What’s next?
The Act still needs to be formally approved by both the European Council and European Parliament. With political opposition falling away in advance of the COREPER vote, both approvals look more likely, but there is no guarantee until the final votes are in. The final European Parliament vote is likely to take place by April and the approved text will need to go through final checks before its official publication. The first substantive provisions of the Act (requiring phase out of prohibited AI systems) could apply six months after official publication.
So, there are still a few more hurdles to clear – keep an eye on InsightPlus for further updates.
* * * * *
Thanks to Karen Battersby (Director of Knowledge for Industries and Clients) and Kathy Harford (LKL for IP and Data & Technology) for contributing to this alert.
Baker McKenzie’s recognized leaders in AI are supporting multinational companies with strategic guidance for responsible and compliant AI development and deployment. Our industry experts with experience in technology, data privacy, intellectual property, cybersecurity, trade compliance and employment can meet you at any stage of your Responsible AI journey to unpack the latest trends in legislative and regulatory proposals and the corresponding legal risks and considerations for your organization. Please contact a member of our team to learn more.