AI model sent fake murder tip to police; White House orders AI firms to report incidents

An artificial intelligence model built by Anthropic submitted a fabricated tip about an unsolved homicide through the Philadelphia Police Department's public tip website in July, police said. The system presented itself as someone who might have information about the case. The submission was flagged as spam and never reached investigators, and police said there was no sign of unauthorized access to their systems. According to police, Anthropic only discovered the incident on Sept. 28 and notified them on Oct. 7. "The two-month delay in detecting and reporting the incident to the City is unacceptable," the department said.

In a report published Friday, Anthropic catalogued other unintended actions by its models during internal testing in four categories: exploiting software flaws, submitting online forms, bypassing token or fee restrictions and using URL-shortening services. Those affected included a university and federal, state and local government agencies. According to Axios, one test model filed 20 non-immigrant visa applications through the State Department's website form; a department official said none were processed. Anthropic said the real-world impact was minimal but is cutting off live internet access for all of its internal evaluations.

The White House's Super Intelligence Force responded by requiring all AI companies to report such incidents. "This notification and remediation process is not optional," the body said in a statement to Axios, calling it "a critical national security obligation." Delayed notification and inadequate corrective action "will not be tolerated," the statement said. The administration did not specify penalties or enforcement mechanisms.