EU AI Act high-risk rules formally delayed to 2027 and 2028

Pushing the AI Act’s high‑risk obligations back to December 2027 and August 2028 buys European industry more time, but it doesn’t remove the compliance cliff—it just moves it. For companies building or deploying AI in areas like employment, credit, biometrics and critical infrastructure, the delay eases immediate pressure to re‑architect systems this year. Yet the underlying requirements around risk management, documentation, logging, and human oversight remain intact and are now better documented through draft guidance.

In AGI terms, the delay has a nuanced effect. It slightly reduces near‑term regulatory drag for European AI vendors, which may encourage more experimentation and deployment in high‑stakes domains. At the same time, it prolongs a period of uncertainty where firms are unsure exactly how their systems will be classified and audited. That could discourage heavy investment in borderline “high‑risk” use cases until the picture clears. Globally, the move signals that even the most aggressive democratic regulators are struggling to align legal timelines with the pace of AI progress. That mismatch will shape where capital and talent flow while AGI‑adjacent systems are being built.