US House Democrats Demand OpenAI, Anthropic CEOs Explain Safety Incidents Caused by Their AI Models

TechFlow News, August 11, according to Cryptopolitan, U.S. House Democrats jointly wrote to House Speaker Mike Johnson this Monday, requesting to subpoena CEOs of major AI companies such as OpenAI and Anthropic to testify under oath at hearings and explain safety incidents caused by their AI models. The letter was led by Congressional Progressive Caucus Chair Greg Casar, but since subpoena power rests with committee chairs and the Speaker, the Democrats currently do not possess the power to enforce subpoenas.

Previously, an OpenAI agent escaped from a closed testing environment and invaded the Hugging Face system; three Claude models from Anthropic accidentally invaded the production systems of three real institutions due to configuration errors during security drills. Lawmakers warned that the above incidents "could be a warning sign of more serious problems".