EU lawmakers back delay to high-risk AI rules
The European Union is delaying key AI obligations for high-risk systems while tightening rules on deepfake and watermarking, highlighting how commercial and political pressure is reshaping the bloc’s landmark AI law.
The European Union has agreed to weaken parts of its flagship AI law, delaying major duties for high-risk systems after pressure from businesses and some governments.
Reuters reported on 7 May that EU governments and European Parliament negotiators reached a provisional deal after nine hours of talks. The package still requires formal approval from member states and the Parliament.
The most important change is timing. Rules for high-risk AI systems linked to biometrics, critical infrastructure and law enforcement will now apply from 2 December 2027, not August 2026. Sky News also reported that AI used in products covered by sector safety rules, including toys, will face a later deadline of August 2028. Machinery will be excluded from the AI Act because other laws already cover it.
Whilst AI supporters say the changes reduce bureaucracy and give European businesses more room to compete with rivals in the US and in Asia, critics say the retreat matters because the AI Act was sold as Europe’s answer to a fast-moving technology that needs democratic limits, not just market incentives.
The deal does include some tighter measures. According to Sky News, systems used to create non-consensual sexually explicit images will be banned from December, and AI-generated content will have to carry watermarks.
GOING FURTHER
EU waters down landmark AI regulation amid industry pressure | SKY NEWS
Artificial Intelligence: Council and Parliament agree to simplify and streamline rules | COUNCIL OF THE EU
Artificial Intelligence Act: delayed application, ban on nudifier apps | EUROPEAN PARLIAMENT
EU reaches tentative deal on simpler AI rules | DW
Sources:
▪ This piece was first published in Europeans TODAY on 8 May 2026.
▪ Cover: Dreamstime/KORWEN.
