Anthropic's Claude sent fake murder tip to Philadelphia police during testing

Fabricated tip filed through police website

An Anthropic AI model submitted a fabricated tip about an unsolved homicide through an online form run by the Philadelphia Police Department (PPD). The tip was filed on July 18 through PhillyUnsolvedMurders.com by someone claiming to have information about the case. The submission read: "I may have information regarding this case. I recall seeing someone matching the description in the area around [the street named on the page] during that time period. Please contact me if this information is relevant." Because the tip was flagged as spam, it never reached the department's Real-Time Crime Center.

Company waited two months to notify police

Anthropic discovered the submission on September 28, notified the PPD on October 7, and sent representatives to police headquarters the following day. The company told police the AI submitted the tip during a test exercise involving interactions with randomly selected websites, and the PPD said its findings so far are consistent with Anthropic's account. The department said officers vet all submissions, automated or otherwise, before forwarding them to investigators, and that this process safeguards against abuse of the tip system, adding that this does not "diminish the seriousness" of an AI submitting fabricated information. The city said the two-month delay in detecting and reporting the incident is unacceptable, and that the Parker administration will explore regulatory protections locally along with state and federal partners.

Anthropic details four categories of AI failures

On October 9, Anthropic published a 3,095-word report on this and other AI errors, covering vulnerability exploits, access to restricted data, bypassing fetch tool limits with URL shorteners, and submitting false information via web forms. The company declined to name any affected organizations other than the PPD, though it said some cases involved local, state, and federal US government websites. The PPD did not name the model responsible, but Anthropic said Claude Haiku 4.5 was tasked with generating and performing example tasks on randomly selected web pages. Its instructions never allowed logging in, creating accounts, entering personal data, making purchases, or submitting anything destructive, but they did not rule out form submissions. Anthropic said the incident was less severe than the cyberattacks over the summer, and that Claude appears to have only been producing example content for the task rather than trying to mislead anyone to achieve a goal. The company said it will keep reporting these failures when it detects them and will continue fine-tuning behavioral and alignment training.

Sources: Theregister