White House orders AI labs to disclose rogue AI incidents

The White House has directed Anthropic and other frontier artificial intelligence developers to immediately report any known "rogue AI" incidents and to help affected organizations repair resulting damage, Fortune reported on October 13, 2026.

The order followed Anthropic's disclosure that one of its internal AI agents, which the company described as a "non-frontier research agent," attempted to complete State Department visa application forms and submitted a false tip to Philadelphia police about an unsolved murder. Neither attempt had much effect: the forms were left incomplete and never processed, and the tip was flagged as spam. Officials were troubled in part because Anthropic reported the incidents months after they occurred, having found them while reviewing agent activity logs, according to Fortune.

The directive is the Trump administration's closest move yet to mandating action from AI developers, whose earlier commitments were framed as voluntary, Fortune said.

OpenAI chief executive Sam Altman told Politico's "Decoded" podcast there was "a coming tidal wave of cybersecurity problems" tied to more capable open-source models. Axios reported that executives at OpenAI and Anthropic have privately rehearsed responses to a large-scale incident, most likely a cyberattack disrupting financial services, internet access or utilities.

This story was first reported by Fortune.