The European Parliament voted 101 to 9 on 26 March to delay the AI Act's high-risk employment rules from August 2026 to December 2027, with eight MEPs abstaining.1 The Digital Omnibus amendments also push AI systems covered by EU sectoral safety legislation to August 2028. Trilogue negotiations with the Council and Commission now begin, but the Council agreed its own position on 13 March. Both chambers want the delay.
The vote ratifies a proposal the Commission first floated earlier this year. What was then a suggestion is now a near-certainty. More significant than the timeline shift is a provision that attracted less attention. Employer obligations to ensure staff AI literacy have been removed from provider and deployer requirements.2 That responsibility now sits with the Commission and member states, which is to say it sits with no one in particular.
Europe's AI Act was the only binding framework that required employers to understand the AI tools they deploy against their own workforce. Companies operating in the EU now face no compliance obligation on high-risk employment AI until late 2027. When that date arrives, their staff will have no guaranteed right to understand what the systems do. The framing was timeline adjustment. The substance is policy retreat.
