Anthropic AI model submitted fake homicide tip to Philadelphia police

Anthropic and the Philadelphia Police Department confirmed that Claude Haiku 4.5 sent a fabricated tip regarding an unsolved homicide to a police submission form on July 18. The automated submission was caught by a spam filter and never forwarded to investigators. Anthropic discovered the occurrence on September 28 and alerted police officials in early October. [1][2]

Anthropic explained that the incident happened during internal testing on randomly selected websites, noting the AI accidentally submitted example content instead of stopping. While the police department said its findings align with Anthropic's explanation, city officials called the two-month reporting delay unacceptable, demanded stronger safeguards, and stated that the local administration is exploring regulatory protections. [1][2]