EU delays implementation of AI Act's rules for high-risk systems until late 2027

By Melike Pala

BRUSSELS (AA) - The European Commission on Wednesday unveiled a digital simplification package, announcing that the implementation of the AI Act's rules for high-risk systems will be delayed until the end of 2027 to ensure key support tools are in place.

EU Commission Executive Vice President Henna Virkkunen told reporters in Brussels that the postponement aims to ensure that "standards, specifications and guidelines" are ready before high-risk obligations take effect, stressing that the move is not a step back from Europe's strict AI safeguards.

The delay is part of a wide-ranging "Digital Omnibus" proposal aimed at simplifying the bloc's growing digital regulatory framework.

The Digital Omnibus, which still requires approval from EU member states and the European Parliament, consolidates and amends multiple laws to ease compliance for companies while streamlining overlapping obligations.

Virkkunen said the extension is necessary to prevent burdensome compliance before essential support tools, such as standards, technical specifications, and guidelines, are finalized.

"This is not about backing down from the rules. It is about ensuring support tools like standards, specifications and guidelines are in place before high-risk rules apply," she noted.

The EU's AI Act, approved earlier this year, sets strict obligations for high-risk AI systems used in areas such as critical infrastructure, employment, education, and law enforcement.

Under the revised timeline, these rules will only become fully enforceable in late 2027, nearly a year later than previously planned.

According to media reports, since the AI Act's adoption, major US and European tech firms have lobbied Brussels to ease rules, while Washington raised concerns over cross-border innovation and AI training.