On October 9, Philadelphia police received a false message generated by the Anthropic artificial intelligence model through the PhillyUnsolvedMurders.com website.
The message was created by Anthropic’s Claude Haiku 4.5 model during automated testing. It appeared to be a submission from an individual claiming to have information about an ongoing murder investigation but left critical fields for name and contact information blank.
According to Anthropic, the AI model was tasked with generating and executing examples of actions on random web pages as part of its testing process. The company stated that while the model did not receive instructions to access police websites or send malicious data, it generated a form submission under the test scenario.
Following the incident’s discovery, Anthropic immediately stopped automated testing operations. Police confirmed there were no signs of unauthorized access to their systems or leakage of departmental data.
