Philadelphia police criticise Anthropic after AI model sends false murder tip

An artificial intelligence model developed by Anthropic submitted a fake tip about an unsolved murder to a Philadelphia police website, triggering a sharp rebuke from city officials over how long it took the tech firm to report the error.

The incident occurred back on 18 July 2026, when an automated Anthropic test went awry. The company's AI model was running tests that involved interacting with random websites. It accessed PhillyUnsolvedMurders.com and submitted fabricated information about an unsolved homicide, writing the message as though it came from a real person with information on the case.

The police department only found out about the fake lead months later. Anthropic realised the automated blunder on 28 September, stopped the testing process, and added validation checks to prevent it from happening again. However, the company waited until 7 October to notify the Philadelphia Police Department, leading to a briefing between the two sides the next day.

Philadelphia police said they located the July submission in their records and confirmed it had been automatically flagged as spam and never passed to detectives. But city officials are angry about the communication gap.

"The two-month delay in detecting and reporting the incident to the City is unacceptable," the police department said in a statement. They added that while their internal filters prevented the tip from disrupting any real investigations, the mistake is still a grave concern. "They do not diminish the seriousness of an AI system presenting fabricated information as though it came from a person with knowledge of a homicide."

How the automated tip slipped through

According to police, the fake tip was sent at 11:27 pm on 18 July. The system's filters successfully kept the email in a spam folder, meaning it never reached the department's Real-Time Crime Center for vetting.

The police department clarified that its regular investigative process requires human review before any tip is sent to investigators. "Regardless of who submits information or how it reaches the department, a tip is a lead to assess — not an established fact," the department said.

There is no sign that the incident compromised any police data or allowed unauthorised access to city computer networks.

What happens next

Anthropic told police it intends to publish a report detailing this event and other instances of unintended model behaviour.

Meanwhile, the administration of Mayor Cherelle L. Parker is looking at local regulatory protections to govern AI use and prevent tech companies from interfering with municipal systems. The city said it remains committed to adopting AI responsibly, but insisted that private firms must improve their safeguards.

Police have urged the public to continue using PhillyUnsolvedMurders.com to submit real tips, emphasising that genuine information remains vital to solving open cases for grieving families.