BRUSSELS — The European Commission is expected to propose delaying a key part of its landmark artificial intelligence rules, according to two Commission officials.

Implementation of the rules on high-risk AI that could endanger people’s safety, health or fundamental rights will be delayed by at least one year under a proposal coming next week, the officials said.

The U.S. administration, U.S. tech companies and lobby groups have put heavy pressure on the European Union not to overregulate artificial intelligence. They have been among the most vocal proponents of a partial “pause” to the EU’s flagship AI laws.

Delaying the rules until August 2027 would mark a shift in the EU’s position in the global artificial intelligence race. It previously wanted to be seen as a global front-runner in regulating the nascent technology but now aims to stay competitive with the U.S. and China.

The proposal from the Commission — to be branded as an amendment to the AI law as part of a wider digital simplification package coming Nov. 19 — would also need sign-off from EU countries and the European Parliament before it is finalized.

By moving to amend the AI Act adopted last year, the EU executive gives in to a growing chorus of EU countries, companies and industry lobby groups.

The critics had demanded a delay, arguing the technical standards that companies could rely on to comply with the high-risk AI requirements were not ready by a summer deadline.

Details of the proposal were presented at a meeting of Commission cabinet specialists on Thursday. The proposal remains open to discussion among the heads of cabinets of the Commission later this week, and at the highest political level among commissioners next week, before being included in the final text of a digital simplification package Nov. 19.

An earlier draft of the package, obtained by POLITICO, also introduced a one-year grace period for fines related to the watermarking of AI-generated content. The timeline on high-risk rules was marked as under consideration.

Other parts of the EU’s AI Act, such as banning certain practices for the most complex and advanced AI models, took effect in February and August this year.

Spokesperson Thomas Regnier said the Commission would not comment on leaks.

“When it comes to potentially delaying the implementation of targeted parts of the AI Act, a reflection is still ongoing within the Commission. Various options are being considered but no formal decision has been taken at this stage,” Regnier said.