EU set to postpone key high-risk AI rules
The European Commission is expected to propose delaying implementation of the EU’s high-risk AI rules by at least one year. According to Commission officials, the postponement responds to concerns that technical standards are not ready and follows pressure from the U.S. administration, U.S. tech companies, and lobby groups.
The European Commission is preparing to delay a key part of the EU’s Artificial Intelligence Act, according to two Commission officials. The proposal, expected next week, would push back the implementation of rules covering high-risk AI systems by at least one year. These provisions apply to technologies that could affect people’s safety, health, or fundamental rights.
The delay comes after significant pressure from the US administration, US tech companies, and industry lobby groups, which have argued that the EU risks overregulating AI. They have been among the most vocal supporters of a temporary pause on parts of the legislation. Postponing the rules until at least August 2027 would represent a shift in the EU’s earlier position, when the bloc sought to be the leading global regulator of AI. The Commission now aims to maintain competitiveness with the United States and China.
The adjustment is expected to appear as an amendment to the existing AI law, forming part of a wider digital simplification package, the Digital Omnibus, set for release on 19 November. Any changes will require approval from EU countries and the European Parliament. The move reflects growing calls from within the EU, including from member states, companies, and lobby groups, who argue that the technical standards needed for compliance will not be ready by the current deadline.
Details of the proposal were shared during a meeting of Commission cabinet specialists. Discussions will continue among heads of cabinets and Commissioners before the final text is issued later in November. A (leaked) draft version of the package, obtained by POLITICO, included a one-year grace period for fines related to watermarking of AI-generated content. The timeline for high-risk rules was marked as still under consideration.
Several parts of the AI Act have already entered into force, including bans on certain practices for advanced AI models, which took effect in February and August of this year.
Asked about the potential delay, Commission spokesperson Thomas Regnier said the institution does not comment on leaks. He added that discussions are ongoing, and while multiple options are being examined, no formal decision has been made.
