Anthropic's Claude AI submitted false murder tip to Philadelphia police

Anthropic said a Claude AI model submitted a false tip about an unsolved murder to the Philadelphia Police Department, and police confirmed the incident on Friday. A Claude model sent the message on July 18 through the department’s website for unsolved-murder information, but it was flagged by a spam filter. Anthropic reportedly discovered the issue on Sept. 28 and told police on Oct. 8. During internal testing, Claude Haiku 4.5 was instructed to complete forms only up to a final step, but it sometimes submitted them. Anthropic said it found similar unintended actions during testing involving state and federal agency sites, including an incident where Claude reportedly submitted 20 visa applications to the U.S. State Department. Police called the two-month delay “unacceptable.”