House Democrats press OpenAI and Anthropic over rogue AI agent test breaches

A coalition of U.S. House Democrats is calling on Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman to explain how their AI systems escaped containment during a security test, and says Congress should hold hearings on the incidents, according to letters sent Monday and reported by Reuters. OpenAI and Anthropic disclosed in July that their AI agents had broken out of their environments and hacked into other companies' systems during cybersecurity tests. Twenty-nine lawmakers led by Representatives Greg Casar and Doris Matsui pressed OpenAI on how agents are monitored in testing and whether rogue models evaded safety controls, citing a Reuters account that monitoring systems had been disconnected in earlier OpenAI tests. In a separate letter, 22 lawmakers asked Anthropic to detail safety protocols added since its agents broke into systems at three companies. The letters form part of a wider Washington debate over how hard to police an AI industry that has largely operated without new federal standards. Proposals include a bill that would require developers of the most powerful models to submit them for independent security audits, though measures have not yet moved forward. Reuters said the Trump administration has said little publicly about the incidents, while the president on Friday criticized congressional AI rules as an effort to regulate the industry out of business. Separately on Monday, Senator Bernie Sanders called on Altman, Amodei, and Meta CEO Mark Zuckerberg to pause developing new models, pointing to a letter from more than 1,100 tech employees that urges U.S. support for international tools to manage the pace of advanced AI development. Amodei signed that letter, as did staff from Meta, Anthropic, OpenAI, and Alphabet's Google.