Overview
- The EU AI Act's core transparency rules became binding on 2 August 2026, giving national supervisors the legal power to monitor compliance and begin enforcement across all member states.
- Any AI system that interacts with people must be clearly identified as AI and include machine‑readable markers such as metadata or watermarks to allow automated detection.
- Operators must disclose synthetic content—AI‑generated or manipulated text, images, audio or video—when no human editorial review takes place, and chatbots must state they are not human at the start of conversations.
- Legacy systems in use before 2 August 2026 have a transition window until 2 December 2026 to implement technical fixes while designated high‑risk AI systems face a later compliance deadline of 2 December 2027 with strict rules on data quality, documentation, monitoring and human oversight.
- Supervisors including Germany's BaFin can inspect AI use in sectors such as banking and insurance and levy steep fines of up to €35 million or 7% of global turnover for certain breaches and up to €15 million or 3% for high‑risk violations.