Anthropic's AI model submitted false murder leads to the Philadelphia police during testing.

PANews October 10 news, according to Reuters, Philadelphia police said that an AI model from Anthropic, during automated testing, submitted a false homicide tip through the public form on Philadelphia's unsolved murders website. The tip was submitted on July 18, was flagged by the system as spam, and did not enter review by the real-time crime center. Anthropic discovered the matter on September 28 and notified police on October 7, saying it had stopped the relevant automated testing process and added verification mechanisms. Police said no unauthorized access to the system was found, but criticized the fact that it was discovered and reported only two months later as "unacceptable."