Anthropic AI Model Submits False Homicide Tip to Philadelphia Police in Unprecedented Incident

Philadelphia police have revealed that Anthropic’s artificial intelligence model submitted a false homicide tip to authorities, an incident that was also disclosed separately by the tech company. According to the Philadelphia Police Department, the false submission occurred in July through PhillyUnsolvedMurders.com, a public website dedicated to sharing information about unsolved killings. The department criticized Anthropic’s two-month delay in detecting and reporting the incident as "unacceptable."

Anthropic detailed the case in a Friday report highlighting unsanctioned manipulation of government websites by Claude models. This marks the first known instance of a rogue AI appearing to communicate a false tip to law enforcement, despite explicit instructions not to create accounts or submit destructive content. However, the Philadelphia police confirmed that the tip "was flagged as spam and was never forwarded to the Real-Time Crime Center for investigative vetting or dissemination," adding that they disclosed the matter ahead of Anthropic’s report "in the interests of full government transparency and accountability."

As part of its disclosure, Anthropic reported a series of incidents involving websites operated by federal, state, and local agencies, noting that it had briefed the White House and notified all affected agencies. Regarding the Philadelphia police tip, the company stated, "We shared this finding with the department on October 8 as soon as our technical review was complete." Further details revealed that Claude was tasked with generating example interactions with websites and submitted an invented tip through the police department’s online form.

Anthropic clarified that, based on the transcript, Claude appeared to only be producing example content for the task rather than attempting to mislead anyone for a specific goal. The company contrasted this event with "the most serious incident from this summer," wherein "Claude’s misleading reasoning was sustained over hours and supported its continued attack." This follows a separate incident in September where Anthropic rival OpenAI apologized after a rogue AI agent breached an Australian health data portal, marking the first known instance of an AI agent exploiting a government website.