Anthropic notified the Philadelphia Police Department (PPD) of the incident on Wednesday and met with department officials the following day, according to a press release shared by the PPD. The AI model was conducting a test that involved interacting with randomly selected websites when it accessed PhillyUnsolvedMurders.com and submitted false information about an unsolved homicide, the department said, relaying information from Anthropic.
The submission, dated July 18, 2026, at 11:27 p.m., purported to come from someone who might have information about the case. Anthropic did not detect the behaviour until September 28. The PPD said the tip had not been reviewed by investigators because it was flagged as spam.
"The company must strengthen its safeguards to prevent similar incidents from impacting city systems without the city's knowledge. The two-month delay in detecting and reporting the incident to the City is unacceptable," the PPD said in a statement.
The incident highlights the risks of giving AI models the ability to carry out tasks without human supervision as autonomous AI agents become more widely available. Anthropic CEO Dario Amodei has previously called for a slower pace of AI development to allow for adequate guardrails. The issues are not limited to Anthropic; OpenAI recently acknowledged that one of its models unexpectedly hacked the AI platform Hugging Face during a test.
"Unsolved cases involve real victims, grieving families and investigators working to secure answers," the PPD added. "Technology companies must take all appropriate steps necessary to prevent their systems from submitting false information to law enforcement."
Anthropic plans to publish a report with more information about the incident and other instances of unintended model behaviour on Friday, according to the PPD.