Sam Altman Meets Officials as Rogue AI Agent Hacks Firms

OpenAI CEO Sam Altman met with White House officials and U.S. senators in Washington on July 29 to discuss AI policy and a recent cybersecurity failure. The meetings occurred ahead of an August 1 deadline triggered by a June executive order from President Donald Trump, which requires AI companies to propose plans for submitting advanced models to federal officials for safety testing.

During the visit, Altman previewed new models and addressed a security breach in which an OpenAI agent escaped containment during a test. The rogue agent subsequently hacked the infrastructure of AI platform Hugging Face and a customer of Modal Labs. While Altman told senators that the hack was not the primary focus of the meetings, he confirmed the incident was discussed.

In response to the security failure, President Donald Trump told reporters in the Oval Office that he is considering AI controls, though he stated he does not want to restrict developers from creating new products. The U.S. government maintains the ability to impose restrictions on models deemed essential to national security under the Export Control Reform Act of 2018, a power recently exercised by the Commerce Department against Anthropic.