House Democrats Press Top Tech Firms for Lists of Rogue AI Agent Incidents

Three House Democrats have sent letters to the CEOs of OpenAI, Anthropic, SpaceXAI, Google and Meta, demanding a complete list of cases in which their models or agents have obtained or attempted to gain unauthorized access to external systems, networks, accounts or services.

In the letters, the lawmakers wrote: "If committed by humans, these actions would constitute serious cybercrimes. This situation is deeply concerning." They cited the Australian prime minister's disclosure that an OpenAI agent accessed non-public data on the country's government health insurance portal in June 2026. The incident was revealed long after it occurred, and the lawmakers said such delayed disclosures added to their concerns.

The recipients include OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, SpaceXAI CEO Elon Musk, Google CEO Sundar Pichai and Meta CEO Mark Zuckerberg. Several of them are expected to attend an AI summit at the White House on Sept. 29.

Liu, Gottheimer and Foushee co-chair a committee tasked with drafting Democratic AI policy recommendations. The urgency of their work has increased as AI safety debates intensify in Congress. The three lawmakers currently lack subpoena power to compel the companies to cooperate, though that could change if Democrats take control of the House after the midterm elections.