Overview
- The EU activated core provisions of the Artificial Intelligence Act that became enforceable on Sunday, August 2, 2026, bringing new transparency duties and oversight of powerful models into operational force.
- Article 50 now requires interactive systems to disclose when users are talking to AI and forces visible labels plus machine‑readable markers on AI‑generated or manipulated images, audio and video to aid automated detection.
- The European Commission’s EU AI Office and national authorities can now request model evaluations before release, inspect technical records, order fixes or market withdrawals, and levy fines up to €15 million or 3% turnover for Article 50 breaches and up to €35 million or 7% for prohibited practices.
- Regulators have set up a 60‑member scientific panel, named advisers, opened secure whistleblower and complaint channels, and are already engaging GPAI providers such as OpenAI and Anthropic after reported unauthorized cyber activity linked to models.
- Major parts of the Act will phase in later with bans on non‑consensual sexually explicit AI and child sexual abuse material from December 2, 2026, and broader high‑risk obligations from December 2, 2027, but legal ambiguities in exemptions and limited enforcement capacity create near‑term challenges and potential transatlantic friction.