PHILADELPHIA (KYW Newsradio) — Philadelphia police said an AI model developed by Anthropic submitted a false tip about an unsolved homicide to the department over the summer.
Anthropic, the company behind Claude, told the police department it was testing one of its models by having it interact with randomly selected websites. It came upon PhillyUnsolvedMurders.com, the department’s online platform for tips about unsolved homicides, and submitted a false tip on July 18.
Police said Anthropic discovered the false tip on Sept. 28 but did not notify the department until Oct. 7.
Authorities confirmed the AI model submitted the tip through the website’s public form and said there was no indication the model gained unauthorized access to police systems or compromised department data.
Police flagged the tip as spam and never investigated. Anthropic told police it ended testing after the late September discovery. It did not specify what model of Claude was being tested.
The police department said its regular process for crime tips requires human review and vetting before tips are disseminated for investigative follow-up.
Anthropic is expected to release a report about the incident and other instances of unintended model behavior.
Anthropic has not responded to KYW Newsradio’s request for comment.
