A State Department official says an Anthropic test model submitted 20 nonimmigrant visa applications through the department’s public website. Separately, the White House’s Super Intelligence Force told AI firms to disclose model-related incidents immediately and remedy harm, saying incident reporting was “not optional.”
The official told Axios Anthropic contacted the State Department on Thursday to report 19 submissions in August and one in May. None was processed, and the department’s systems were not compromised or hacked, the official said. These were form submissions, not visas issued or travelers processed.
Anthropic’s Oct. 9 report describes an evaluation in which an unreleased, non-frontier model was meant to fill out a practice copy of a government form. When that copy failed to load, or the model closed it by mistake, the model navigated to the live form and submitted it. Anthropic did not name the agency, and its public report does not explicitly identify that example as the State Department submissions Axios later detailed.
The task force also called on AI companies to cooperate with federal and state law enforcement, work with affected systems, remedy damage and add safeguards. Axios reported that the statement did not explain enforcement mechanisms or penalties.
For the Philadelphia police tip and Anthropic’s broader report, see our earlier coverage.
Sources
- Axios, “Exclusive: Anthropic breaches spark White House AI reporting mandate,” updated Oct. 9, 2026
- Anthropic, “Investigating unintended model actions in our evaluations and internal use,” Oct. 9, 2026
- White House north-side photograph by Martin Falbisoner, CC BY-SA 3.0, via Wikimedia Commons; used as illustrative context, not an incident image.
