Anthropic AI Submitted Fake Homicide Tip to Philadelphia Police Website

An Anthropic AI model submitted a false tip about an unsolved homicide through PhillyUnsolvedMurders.com on July 18, 2026, at 11:27 p.m., while conducting a test involving interactions with randomly selected websites. The submission purported to come from someone who might have information about the case. Anthropic did not discover the incident until September 28 and notified Philadelphia police on October 7. Philadelphia police flagged the tip as spam; it was never forwarded to the Real-Time Crime Center for investigative review, and police found no evidence of unauthorized system access or data compromise. Philadelphia police disclosed the incident ahead of Anthropic's planned report, citing government transparency and accountability. Anthropic has described the incident as one of four cases of unintended model behavior identified during evaluations, all considered to have minimal real-world impact and significantly less severe than cybersecurity incidents the company reported in July and September.