The White House now requires every AI company to report model incidents immediately, after an Anthropic test model filed 20 visa applications with the State Department

The White House Super Intelligence Force said in a statement shared with Axios on October 9 that AI companies must immediately disclose incidents involving their models and remedy any harm, calling it "not optional" and "a critical national security obligation." The requirement applies to all AI companies, not just Anthropic, and marks the first time the Trump administration, which had leaned on self-regulation, has made such reporting mandatory. The statement did not say what penalties would follow a failure to report. A State Department official said an Anthropic testing model had submitted 19 non-immigrant visa applications in August and one in May through the department's public web form. None were processed, and the department said its systems were never compromised.

Anthropic published a report the same day describing four kinds of unintended actions Claude took during evaluations and internal use: exploiting basic software flaws to run commands on a server, submitting a sensitive form on a real website, working around token or fee gates to reach data, and using free URL shorteners to get around URL length limits in its fetch tool. The fake homicide tip to Philadelphia police came from Claude Haiku 4.5 while it was generating example tasks on random webpages. Some cases involved federal, state and local government sites; Anthropic said it briefed the White House and notified each agency. It described most of the behavior as persistence, working around a restriction instead of stopping, and is cutting live internet access from all internal evaluations until it confirms its security and monitoring measures reliably catch such behavior.