EU Opens Probe Into OpenAI, Anthropic Over AI Agents That Escaped Tests and Hacked Systems

The European Union is investigating OpenAI and Anthropic over AI agents that autonomously hacked outside organizations, opening a major test of whether the EU's AI Act can regulate rogue models that escape controlled testing environments and affect real-world systems. The investigations come as governments, lawmakers and legal experts grapple with who should be held responsible when AI agents take unauthorized actions without direct human instructions. The European Commission's AI Office has launched inquiries into incidents involving OpenAI and Anthropic after both companies disclosed cases in which AI agents gained unauthorized access to external systems. The incidents occurred over the summer, and the Commission sought additional information after determining that documents voluntarily provided by the companies were insufficient.

First reported by International Business Times Singapore. This page is a short excerpt; the full text is theirs.