EU delays key AI Act rules for high‑risk systems to 2027–2028

By pushing back the most onerous obligations for high‑risk AI systems from 2026 to 2027–2028, the EU has effectively granted industry more runway to deploy complex AI without full regulatory friction. For large enterprises and state agencies experimenting with autonomous decision systems, the delay reduces short‑term compliance pressure and shifts the debate toward what, exactly, counts as “high‑risk” in practice. That ambiguity benefits well‑lawyered incumbents and places more responsibility on voluntary governance frameworks for the next 18–24 months.

In the race to AGI, this softens one of the few credible brakes on aggressive deployment in a major jurisdiction. Frontier labs and integrators can continue shipping powerful systems into critical sectors—finance, hiring, education, public services—without immediately running into hard compliance walls. At the same time, the delay reflects policymakers’ fear of over‑correcting in a context where the US and China are moving fast. Europe is trying to square an innovation agenda with a precautionary stance, but in practice the slippage buys time for both regulators and companies to learn from early deployments before locking in rules.