Commission to delay AI Act high-risk provisions to 2027

The European Commission is preparing to propose a delay of at least one year to the high-risk provisions of the AI Act, according to Commission officials. The targeted postponement would push implementation to August 2027 and form part of a broader “digital simplification” package expected on November 19. The amendment would require approval by EU Member States and the European Parliament.

The shift responds to sustained pressure from the U.S. administration, U.S. tech firms, and industry groups, as well as a growing chorus within the EU arguing that technical standards for high-risk AI compliance are not yet ready. The Commission’s move signals a recalibration from its early leadership stance toward competitiveness with the U.S. and China, while attempting to preserve the Act’s core safeguards.

An earlier draft of the package reportedly included a one-year grace period for penalties tied to watermarking AI-generated content, while the high-risk timeline remained under consideration. Other AI Act provisions have already entered into force this year, including bans on certain practices for advanced AI systems, maintaining the overall architecture of the regulation.

Commission spokesperson Thomas Regnier indicated no formal decision has been taken and declined to comment on leaks. The proposal has been presented to cabinet specialists and will be discussed among heads of cabinets and commissioners prior to inclusion in the final text. If adopted, the delay would provide industry and standard-setters time to finalize harmonized technical standards without undermining the Act’s existing prohibitions.