AI Act on track to be formally adopted ahead of the European elections
Written By
Groundbreaking new legislation to regulate artificial intelligence systems on the EU market is now on course to be officially adopted by the European Parliament in April and thereafter by the Council, ahead of the European elections in June. To recall, on 8 December 2023, EU institutions reached a provisional political agreement on the EU Artificial Intelligence Act (AI Act), which nevertheless left many open provisions regarding scope, enforcement, governance, prohibitions, general purpose AI (GPAI) and high-risk applications. Throughout January, several technical meetings took place between the European Commission, Council and the European Parliament to hash out agreements on these significant issues.
A finalised text of the Act, which is presented as taking a risk-based approach to AI governance, was distributed to EU Member States on 22 January before the official publication of the text on 26 January. Concerns expressed that Germany, France, Italy and Austria might vote against the file since there were still reservations about the final text. For Germany and France, the main sticking point appeared to be the approach towards general-purpose AI models, amid fears that overly restrictive regulation could jeopardise European innovation in AI.
However, the current Belgian Presidency of the Council had little appetite for further discussions due to the tight timetable to get the regulation adopted within this European Commission and European Parliament mandate, and therefore resisted significant changes and presented a ‘take-it-or-leave-it’ scenario. A subsequent vote to endorse the text at the Committee of Permanent Representatives (COREPER), comprising deputy ambassadors from the 27 EU Member States, passed unanimously on 2 February.
Decision adopted on AI Office
On 24 January 2024, the European Commission published a decision establishing the EU Artificial Intelligence Office, which was pushed through quickly to get ahead of adoption of the AI Act. The AI Office will be part of the Commission’s Directorate-General for Communication Networks, Content and Technology (DG-CNECT). Its primary role will be to provide guidance to the Commission regarding the implementation and application of the incoming AI framework.
Among the key tasks of the future AI Office are the following:
- Developing tools, methodologies and benchmarks for evaluating capabilities of general-purpose AI models (GPAI), in particular those with systemic risks;
- Monitoring the implementation and application of rules on GPAI models and systems, in particular where the model and the system are developed by the same provider;
- Monitoring the emergence of unforeseen risks stemming from GPAI models, including by responding to alerts from the scientific panel;
- Investigating possible infringements of rules on GPAI models and systems, including by collecting complaints and alerts, assisting in the preparation of decisions of the Commission and conducting evaluations pursuant to the forthcoming Regulation;
- Supporting the implementation of rules on prohibited AI practices and high-risk AI systems in coordination with relevant bodies responsible under sectoral legislation;
- Contributing to a strategic and coherent EU approach to international initiatives on AI;
- Monitoring the evolution of AI markets and technologies;
- Assisting in the preparation of relevant Commission Decisions, implementing and delegated acts and guidelines to support the practical implementation of the forthcoming Regulation; and
- Contributing to the establishment and operation of AI regulatory sandboxes.
The decision to establish the EU AI Office will come into effect on 21 February 2024.
Next Steps
The consolidated AI Act text adopted by the EU deputy ambassadors (COREPER) will now be subject to review by lawyer-linguists. The European Parliament is expected to vote on the text at committee level in mid-February, followed by a plenary vote of the whole European Parliament to adopt the text formally in April. Thereafter, the text also needs to be formally adopted by the Council at Minister level.
Once the final text has been published in the Official Journal of the EU, it will enter into force 20 days later, triggering the gradual application of the rules. While the AI Act will apply 24 months after it enters into force, the provisions regarding prohibitions will apply six months after the Regulation’s entry into force. Requirements for GPAI will apply 12 months after the Regulation’s entry into force. Some requirements for high-risk AI systems will apply 36 months after the AI Act’s entry into force.
For more information please contact Paula Alexe or Francine Cunningham.
SIGN UP TO OUR CONNECTED NEWSLETTER FOR A MONTHLY ROUND-UP FROM OUR REGULATORY & PUBLIC AFFAIRS TEAM
