EU’s AI Act moves into enforcement: regulators gain power to fine violators up to 7% of global revenue

The European Union’s landmark AI Act has entered a new phase: as of 2 August 2026, the European Commission and national regulators have full powers to investigate and penalize AI providers that break the law. Authorities can now impose fines of up to 7% of a company’s worldwide annual turnover for the gravest violations, such as deploying banned “unacceptable‑risk” systems or ignoring key transparency duties for general‑purpose AI models. The list of prohibited practices includes certain manipulative systems that may cause serious physical or psychological harm, as well as non‑consensual deepfakes and AI‑generated child sexual abuse material.

Many of the most detailed obligations for “high‑risk” AI—covering applications in areas like healthcare, finance, critical infrastructure, and law enforcement—have been postponed to late 2027 and beyond, after industry groups warned they were not ready. For now, the focus is on transparency: general‑purpose model providers must publish technical information, assess systemic risks, and support downstream users’ compliance, while deployers of AI‑generated content face stricter labelling rules. Because the AI Act applies to any provider that offers AI services in the EU market, it will effectively bind US and Chinese tech giants as well. Legal scholars increasingly see it as an “AI‑era GDPR,” likely to shape global norms as other jurisdictions copy or adapt its approach.