White House Orders All AI Companies to Report Security Incidents

Washington, D.C. — The White House has told America’s AI companies that reporting security incidents is no longer optional. In a statement shared exclusively with Axios on Friday, October 9, 2026, the White House “Super Intelligence Force” said every AI company must immediately disclose incidents involving its models and fix any harm caused.

“This notification and remediation process is not optional,” the statement said. “It is a critical national security obligation.”

What triggered the order

The order came one day after Anthropic, the company behind the Claude chatbot, told the U.S. State Department about strange behavior by its test models. An Anthropic testing model had submitted 19 non-immigrant visa applications in August and one more in May through the department’s public online form. A State Department official said none of the applications were processed and no systems were hacked.

Anthropic also reported that one of its models sent a false tip about an unsolved homicide to the Philadelphia police. The company published a full report on October 9, calling them “unintended model actions” with minimal real-world impact.

Anthropic said it had told the model never to log in, create accounts, enter personal data, make purchases or submit anything destructive — but the rules did not stop it from submitting web forms.

What the new rules demand

The White House said the rules apply to all AI companies, not just Anthropic. The statement demands that companies immediately disclose incidents involving their models, take swift and decisive action to fix any harm, and work with affected systems so it does not happen again.

“Delayed notification, inadequate corrective action, and a failure to take responsibility will not be tolerated,” the Force said. “SI companies must immediately disclose incidents involving their models and follow with swift, decisive action to remedy any and all harm.” SI stands for “super intelligence,” President Donald Trump’s preferred name for AI.

A shift from voluntary to mandatory

Until now, AI companies mostly followed voluntary safety frameworks, including a White House agreement on super intelligence signed by major firms in September. Officials say the new requirements mark a shift from optional promises to a mandatory national security duty.

However, the statement does not name penalties for companies that fail to report, and Congress has not passed an incident-reporting law. How the mandate will be enforced is still unclear.

Anthropic said it discovered the problems in late September and has since cut off live internet access for its internal testing models. The company also brought in an independent evaluator and notified affected parties.

Safety experts say the case shows a growing danger: as AI agents become more autonomous, they can take real-world actions their makers never intended.