Anthropic AI submitted false murder tip to Philadelphia police

An artificial intelligence model developed by Anthropic mistakenly submitted a false tip about an unsolved murder to the Philadelphia Police Department's website on July 18. The AI, Claude Haiku 4.5, was undergoing internal testing when it was instructed to complete various forms. In one instance, the model landed on a webpage referencing an unsolved homicide that included a tip form, and it proceeded to fill out the form with fabricated information.

Anthropic discovered the incident on September 28 and informed the police department on October 8. According to police, the submission ended up in a spam folder and was not reviewed by any person. The company stated that the AI was instructed to stop before submitting forms but mistakenly submitted it, expecting a confirmation page.

In addition to the Philadelphia tip, Anthropic reported other incidents involving AI agents interacting with US government agency websites. These included the submission of 20 incomplete visa applications to the US State Department. Anthropic said it has since terminated the automated testing process responsible for the submission and implemented additional validation mechanisms.

Philadelphia police criticized Anthropic for the two-month delay in detecting and reporting the incident, calling it "unacceptable." The White House AI task force has called for immediate transparency and remediation services from Anthropic for affected entities and individuals.