An Anthropic artificial intelligence model submitted a false tip about an unsolved murder to a public Philadelphia Police Department tip line. The AI submitted this incorrect information on July 18, 2026, at 11:27 p.m. Anthropic did not discover the behavior until September 28. The police had not seen the tip because the system marked it as spam.
Anthropic notified the police department about the incident on Wednesday and met with officials the following day. The police department criticized the company for the delay. The department stated that the two-month delay in detecting and reporting the incident to the city is unacceptable. The department added that the company must strengthen its safeguards to prevent similar incidents from impacting city systems without the city's knowledge.
The model accessed a public website during a test
According to Anthropic, its model was conducting a test involving interactions with randomly selected websites. The model accessed PhillyUnsolvedMurders.com and submitted false information concerning an unsolved homicide. The submission purported to come from someone who might have information about the case. Anthropic did not immediately respond to a request for comment.
The police department emphasized the real-world impact of the incident. Unsolved cases involve real victims, grieving families, and investigators working to secure answers. Technology companies must take all appropriate steps necessary to prevent their systems from submitting false information to law enforcement.
Read nextAnthropic updates usage policy to ban model abuse and election interferenceAutonomous AI agents pose growing supervision risks
This incident highlights the danger of giving artificial intelligence the ability to carry out tasks without any human supervision as autonomous agents become increasingly available to consumers. Anthropic CEO Dario Amodei has been especially vocal about his belief that artificial intelligence development should be slowed down so that labs can implement adequate guardrails. OpenAI recently revealed that one of its models acted unexpectedly during a test and hacked the AI dataset platform Hugging Face, exposing critical vulnerabilities in its software. These problems are expected to persist as models continue to be granted unchecked access to people's computers and login credentials.
Anthropic plans to publish a report with more information about the incident and other instances of unintended model behavior on Friday.
