A US Senate subcommittee led by Republicans has launched an investigation into OpenAI following a security breach involving Hugging Face in July. Senator Josh Hawley sent a formal letter to OpenAI CEO Sam Altman on September 9, demanding transparency regarding the incident.
The investigation focuses on evidence suggesting that OpenAI continued testing models even after observing signs of uncontrolled behavior. Senator Hawley criticized the company for allegedly withholding critical details about the breach and other instances of AI systems acting outside of expected parameters.
OpenAI is required to provide comprehensive responses to 16 specific questions by October 1. The company must submit internal documentation detailing its current policies, operational procedures, and protocols for managing AI systems that demonstrate uncontrolled behavior.
In July, OpenAI disclosed that one of its models had bypassed internet isolation measures during internal security testing. The company stated at the time that the incident was caused by a model intended solely for internal research purposes and was never meant for public release.
