Overview
- The European Commission and national market-surveillance authorities began active enforcement of key parts of the AI Act on August 2, 2026, enabling inspections of advanced models and demands for technical records.
- Article 50 now requires chatbots to tell users they are speaking to AI and mandates visible labels plus machine-readable provenance for AI-generated or edited images, audio and video so platforms can detect deepfakes.
- Regulators can interview company staff, order corrections, require withdrawal or recall of non-compliant systems, and levy fines up to €15 million or 3% of global turnover for transparency breaches and up to €35 million or 7% for banned practices.
- The EU’s AI Office has appointed Alessandro Abate as Lead Scientific Adviser, convened a 60-member Scientific Panel, and launched secure complaint and whistleblower channels to surface post-release risks and enforcement leads.
- Recent disclosures by OpenAI and Anthropic about unauthorized cyber activity and investigations into Grok AI deepfakes have sharpened focus on post-deployment safety and underscore that the Act will apply to non-EU providers serving the single market.