EU starts enforcing AI Act transparency rules, with fines up to €15M or 3% of revenue
The European Commission began enforcing the AI Act's transparency obligations on 2 August 2026: deepfakes, emotion-recognition and biometric categorisation output, and unreviewed public-interest text must carry visible labels plus machine-readable marks, and chatbots, agents, and avatars must disclose they are not human. Enforcement is split across national market surveillance authorities, the European AI Office, and the European Data Protection Supervisor, with fines reaching €15 million or 3% of global turnover. Provenance marking is now a shipping requirement for any AI product with EU users, not a roadmap item.
Source: commission.europa.eu ↗
Transparency obligations will help people recognise when they are interacting with AI or are exposed to AI-generated content, so they can make informed decisions and better protect themselves from misinformation or deception.
Why this matters
- → Concrete enforcement: €15M fines + 3% revenue penalties now active, not future threat.
- → Shipping requirement, not roadmap — AI provenance marking is mandatory for EU market access.
- → Labeling standards clarify what counts as AI-generated; reduces user deception surface.