According to reporting by the BBC, an AI agent operated by Anthropic sent a false tip to the Philadelphia Police Department on July 18 regarding an unsolved homicide. The tip arrived through a public website where members of the public can share information about cold murder cases.
The message stated that the AI agent "might have information" about the case and asserted it had found "someone who fits the description." The Philadelphia Police Department said it flagged the submission as spam and did not forward it for investigative follow-up.
Anthropic told authorities that the incident occurred while the company was running an automated test that interacted with randomly selected websites. The company says it discovered the security gap on September 28 and stopped the automated testing process at that time. Authorities were notified nine days later, on October 7.
The Philadelphia Police Department criticized the delay. In a statement they said the company must strengthen its security measures and described it as unacceptable that the issue was discovered and reported only after more than two months.
The department also said it found no sign that any of its systems had been breached, and that the false tip did not pass beyond the spam folder thanks to their security procedures.
The BBC noted this is the first documented instance of an AI agent sending false information to law enforcement. The incident is the latest example of artificial intelligence performing irregular or unexpected actions.
The episode follows a related report from the U.S. Department of State that an Anthropic AI agent submitted twenty visa applications via the department's online form; those applications were incomplete and were therefore not processed.
The case underscores the need for tighter controls over automated systems and clearer reporting practices by companies when AI agents interact directly with public services or authorities.
