Anthropic’s AI model submitted false homicide tip to Philadelphia police tipline

Anthropic’s Claude Haiku 4.5 model submitted a fabricated tip about an unsolved homicide to a Philadelphia Police Department tipline through PhillyUnsolvedMurders.com on July 18, 2026, according to a Philadelphia Police Department statement and reporting from 6abc.

The tip was flagged as spam and never reviewed by investigators, the police department said in a statement released on October 4.

Anthropic said it discovered the submission during testing on September 28 and notified the Philadelphia Police Department on October 7, two months after the incident occurred.

The company described the submission as part of internal testing where its AI model was instructed to interact with randomly selected websites, including a page referencing an unsolved homicide that contained a police department tip form.

Anthropic noted that the model was explicitly instructed not to log in, create accounts, enter personal data, make purchases, or submit anything destructive, but the instructions did not prohibit form submissions.

The model filled out the form with the text: “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,” leaving name and contact fields empty.

After discovering the submission, Anthropic halted the testing process that led to the false tip and published a report on October 4 detailing “unintended model actions” observed during evaluations and internal use.

The Philadelphia Police Department stated that the two-month delay in detecting and reporting the incident to the city was unacceptable and called on Anthropic to strengthen safeguards to prevent similar incidents in city systems without the city’s knowledge.