EU adopts new AI rules, postpones high-risk obligations

The European Union on Monday adopted new rules aimed at simplifying its artificial intelligence (AI) framework, including a postponement of obligations for high-risk AI systems to ease compliance burdens on businesses and national authorities.

The legislation, approved by the European Council as part of the bloc's "Omnibus VII" simplification package, delays the implementation of key provisions of the AI Act that were originally due to take effect on Aug. 2, according to a statement.

Under the revised timeline, rules for stand-alone high-risk AI systems will apply from Dec. 2, 2027, while requirements for high-risk AI systems embedded in products will enter into force on Aug. 2, 2028.

EU lawmakers fast-tracked the measure amid concerns from industry and member states over the readiness of companies and regulators to comply with the new obligations.

The regulation forms part of the EU's broader effort to simplify legislation and reduce administrative burdens while maintaining safeguards for the use of artificial intelligence.

Alongside the delayed implementation schedule, the new law introduces a ban on AI practices involving the generation of non-consensual sexual and intimate content as well as child sexual abuse material.

AI systems capable of generating nude images of real individuals or digitally removing clothing from existing photographs to reveal intimate body parts will be prohibited from December of this year.

The legislation also extends the deadline for member states to establish AI regulatory sandboxes until Aug. 2027, giving national authorities additional time to set up testing environments for innovative AI technologies.