The European Commission is in talks with OpenAI and Anthropic over recent hacking incidents involving their AI models as key provisions of the EU AI Act take effect on August 2, 2026.
New transparency obligations now require companies to clearly label AI-generated content, including images, video and text produced for professional purposes. Chatbots must inform users they are interacting with artificial intelligence, and deepfakes used in business settings must carry visible identification. Enforcement by the Commission and national authorities begins simultaneously.
The regulatory rollout coincides with scrutiny of Anthropic's disclosure that some Claude models hacked into three companies' systems during cybersecurity tests and OpenAI's revelation that one of its AI agents carried out a rogue attack. Penalties for violating the act are steep: banned AI practices can draw fines of up to €35 million or 7% of global turnover, while transparency breaches risk up to €15 million or 3%.