European Commission proposes delaying full implementation of AI Act to 2027

The European Commission is today unveiling its "Digital Omnibus", a plan meant to simplify the EU's sweeping digital regulations.

Among the legislation it covers is the AI Act, which was adopted last year. The legislation is being implemented only gradually – but with the Omnibus, the Commission has officially announced that some provisions on the riskiest iterations of the technology are set to be delayed.

In particular, the Ominbus offers more time for businesses and organisations deploying "high-risk" AI technologies, which are used for purposes such as analysing CVs, evaluating school exams or assessing loan applications, will now not come under the full weight of the bill's provisions until December 2027, more than a year after the original date of August 2026.

This could have a major concrete effect. If the Digital Omnibus is adopted, AI models will be able to use previously restricted data to make decisions over access to essential financial services.

Peter Norwood, Senior Research and advocacy officer at Finance Watch said that this amounts to a "deregulate to accelerate" AI strategy, and that consumers will ultimately pay the price for it.

“Under these proposals, a person could be denied a loan because of a biased AI model, or charged higher insurance premiums based on predicted health status, all without their knowledge or consent,” he said.

Why this delay?

The Commission has chalked up the postponement to a lack of implementation by member states and companies' need for time to adapt to the complex new rules.

Big tech lobby group CCIA, which counts among its members Amazon, Apple, Google or Uber, welcomed the Omnibus's proposed delay, but called for "bolder" and "clearer" actions.

"The Omnibus misses critical opportunities to raise the outdated compute threshold for identifying AI models which pose a ‘systemic risk’, and fails to fix problematic wording on the extraterritoriality of copyright provisions, which conflicts with EU and international principles," the group wrote in a statement.

Some critics of the Omnibus, however, say the pause goes too far.

“Consumers were promised simplification to support the European economy, and yet the Commission’s proposal can only be read as deregulation almost to the exclusive benefit of Big Tech," said Agustín Reyna, Director General of European Consumer Organisation BEUC.

"Instead of cutting down on consumer rights, the European legislator should focus on making compliance easier to the benefit of European companies and consumers alike."

While France and Germany have already welcomed the AI Act's provisions, many critical AI experts regret that numerous EU member states didn’t put the required enforcement structures in place to comply with the legislation on its original timeline.

“Many member states missed the August 2025 deadline to designate competent authorities,” said Hanane Taidi, Director General of TIC Council, which represents the independent companies responsible for assessing compliance with the AI Act.

“Until these national structures are in place, Conformity Assessment Bodies cannot be notified, and the system cannot fully function.”

The next hurdle

The adoption of the Omnibus might not be easy. The proposed changes to the EU's AI rules will require a controversial modification to the General Data Protection Regulation (GDPR), and many MEPs spanning the left and centre of the political spectrum have already announced their opposition.

Francine Cunningham, Regulatory and Public Affairs Director at the law firm Bird & Bird, told Euronews: “We can expect some resistance ahead since the omnibus process bypasses the usual impact assessments and consultations required for legislative proposals.

"Some of the laws being amended were only recently adopted and, in some cases, haven’t even been fully implemented yet. MEPs may therefore be reluctant to change the laws they only recently debated and adopted."

However, the rules still need to be in place before the deadline of August 2026 – the original date for the implementation of "high-risk" AI rules.