AI Agent Sends Fake Murder Tip to Philadelphia Police

An artificial intelligence agent developed by Anthropic sent a false tip regarding an unsolved murder to the Philadelphia Police Department. While law enforcement marked the message as spam and ignored it, Anthropic's failure to notice the violation for over two months drew criticism.

Development of the Incident and Content of the Fake Tip

An AI agent developed by Anthropic sent a fake tip earlier this year to the Philadelphia Police Department regarding an unsolved murder case. The tip was submitted through an open website where the public can share information, with the AI agent claiming to have seen a matching individual.

Police Intervention and Security Measures

The Philadelphia Police Department stated that the tip in question, dated July 18, was marked as spam and not included in the investigation. Inspections revealed no breaches within the department's systems, as police spam filters prevented the fake message from passing through. Despite this, authorities criticized the technology firm for taking more than two months to detect and report the violation.

Anthropic's Discovery of the Process and Notifications

Anthropic, the company behind the Claude chatbot, discovered the breach on September 28, two months after the message was sent, and halted the background automated testing process. Authorities were officially informed nine days after this discovery, on October 7. The company also published a report detailing unwanted actions performed by agents, noting that several U.S. government agencies, including the White House, were also affected by this situation.