Newsroom

Europe Just Moved the Goalposts on AI Compliance. Here Is What It Means

On 7 May 2026, EU lawmakers reached a provisional agreement to revise the AI Act's enforcement timeline.

Newsroom 7 May 2026

On 7 May 2026, EU lawmakers reached a provisional agreement to revise the AI Act's enforcement timeline. The headline change: the deadline for high-risk AI system obligations, covering credit decisions, employment tools, critical infrastructure, and essential services, has shifted from August 2026 to December 2027, a 16-month extension.

For many enterprises this will feel like relief. It should not.

The compliance requirements have not changed. Technical documentation, automatic event logging, explainability for human oversight, and full decision traceability are still the destination. The extension simply means organisations have more time to reach it, and less excuse not to get it right.

The enterprises that treat this as breathing room rather than runway will face the same structural problem in late 2027 that they were facing now: AI systems built on infrastructure that was never designed to prove its own reasoning. Logging is not provenance. Auditability is not the same as verifiability.

The window is open. What you build inside it is the question.