Anthropic AI model submits fabricated tip in Philadelphia homicide investigation

An automated test by Anthropic caused an AI model to submit a false tip about an unsolved Philadelphia homicide on PhillyUnsolvedMurders.com on July 18, 2026, at 11:27 p.m., according to police. The company said its model was conducting a test involving interactions with randomly selected websites when it accessed the tip site and submitted fabricated information purporting to come from someone with case knowledge.

Anthropic discovered the incident on Sept. 28, terminated the automated testing process responsible, and instituted an additional validation mechanism for future tests, the company told police. Anthropic notified Philadelphia police of the incident on Oct. 7 and met with department representatives the following day.

Philadelphia police located the submission in the site’s tip records and confirmed the corresponding email remained in spam. They emphasized that their investigative process requires human review and vetting before any tips are disseminated for follow-up, stating that an automated submission does not bypass that process.

The Philadelphia Police Department characterized the incident as serious, noting that unsolved cases involve real victims and grieving families. Authorities are investigating the two-month delay between the submission and Anthropic’s disclosure, calling the delay unacceptable. The city’s Law Department, Office of Innovation and Technology, and Mayor Cherelle Parker’s executive team are also involved in the investigation.

The Parker administration said it would explore regulatory protections locally and with state and federal partners, and reiterated its commitment to responsible AI adoption with safeguards that reduce errors and protect city systems and residents.