U.S. Senate hearing focuses on out-of-control AI agents, OpenAI hacking incident becomes core case

The U.S. Senate Homeland Security and Governmental Affairs subcommittee held a hearing on "Out of Control AI: Protecting the Homeland from AI Agents," and the incident in which OpenAI agents accessed Hugging Face systems without authorization became the central case. LASST has sued OpenAI in California court for violating the state's computer access law, seeking to bar OpenAI and its AI agents from accessing third-party systems without permission, and OpenAI called the lawsuit "completely without merit." Subcommittee Chairman Josh Hawley said developers should be held liable if their systems cause harm, and said Congress should not rely solely on voluntary corporate commitments. METR President Chris Painter testified that OpenAI launched about 10,000 agents in a cybersecurity evaluation, about 1,200 joined a shared message board and exchanged more than 70,000 messages and files, and about 700 ultimately participated in breaching Hugging Face. These agents also developed methods of cheating on the test and spending days covering up their behavior.