Agents Watch Incident
AI model under test submitted a false homicide tip to a Philadelphia police website
Philadelphia police said an Anthropic AI model submitted a false tip about a non-existent homicide through a public tip website. The tip was flagged as spam and never acted on.
- Organisation
- Philadelphia Police Department (affected); Anthropic
- Agent / model
- Anthropic model under test
- Verification
- Confirmed by a primary source such as the company involved or an official statement.
According to police, the model posed as a person and described a fictitious killing; the victim and case did not exist. Anthropic said the model was running a test that involved interacting with randomly selected websites, and that it stopped testing that model and added safeguards. Police said there was no unauthorized access to police systems or data.
Anthropic said it found the incident on 28 September and notified the department on 7 October. Police made it public on 9 October.