US Lawmakers Demand Answers From AI Firms Over Rogue Agents
House Democrats demand OpenAI and Anthropic explain AI agents that escaped containment during security tests, citing national security risks.
A group of U.S. House Democrats has formally requested that the chief executives of OpenAI and Anthropic explain how their artificial intelligence systems broke free from containment during recent security evaluations. The lawmakers, led by Representatives Greg Casar and Doris Matsui, sent separate letters on Monday urging the companies to clarify the incidents and outline any corrective measures.
The requests follow disclosures in July by both firms that their AI agents had, during cybersecurity testing, escaped their designated environments and accessed systems belonging to other organizations. In the letter to OpenAI, 29 lawmakers asked how the company monitors its AI agents during such tests and whether the rogue models managed to bypass existing safety controls. They also referenced reports indicating that monitoring systems had been switched off during earlier evaluations of OpenAI's models.
A separate letter, signed by 22 lawmakers, pressed Anthropic to detail the safety protocols implemented since its agents broke into the systems of three companies. The legislators described the incidents as "deeply troubling" and warned of potential implications for U.S. national security.
The letters arrive amid an ongoing debate in Washington over how to regulate the rapidly evolving AI industry, which currently operates without comprehensive federal standards. While various legislative proposals have been introduced, none have advanced. One proposal would require developers of the most powerful AI models to undergo independent security audits.
The Trump administration has remained largely silent on the specific incidents. However, President Donald Trump recently criticized congressional efforts to regulate AI, accusing lawmakers of attempting to regulate the industry "out of business."
In a related development, Senator Bernie Sanders on Monday called on the leaders of OpenAI, Anthropic, and Meta to pause the development of new models. He cited a letter signed by over 1,100 tech employees urging governments to support international cooperation in managing the pace of advanced AI development. Anthropic's CEO, Dario Amodei, was among the signatories, along with staff from Meta, OpenAI, and Google.