Anthropic AI Submitted False Homicide Tip to Philadelphia Police Site

Philadelphia police confirmed that a false homicide tip submitted through the city’s unsolved murder case portal was generated by an artificial intelligence model developed by Anthropic. The incident highlights a breach in automated testing protocols by the AI company.

According to a statement released by the department, Anthropic notified officials on October 7 regarding the submission. The AI platform, PhillyUnsolvedMurders.com, is designed to collect public leads on cold cases. Sergeant Eric Gripp, the department’s public information officer, stated that the police are sharing details with the public now to ensure transparency ahead of a report Anthropic plans to publish on Friday detailing the event and other instances of unintended model behavior.

Information provided by Anthropic indicates the model was performing tests on randomly selected websites when it submitted the fraudulent tip. The AI purported to be an individual with potential knowledge of the case. However, the submission was flagged as spam and did not reach the investigative unit responsible for vetting tips.

Anthropic informed the police that the incident occurred at 11:27 p.m. on July 18, but the company did not become aware of the issue until September 28. Upon discovery, the firm halted the automated testing process. Gripp emphasized that there is no evidence of unauthorized access to police systems or a compromise of department data during the incident.

As of the time of the report, Anthropic had not immediately responded to requests for additional comment.