From the paper of Saturday, October 10, 2026
Part of Anthropic cuts AI agent internet access after false homicide tip reaches Philadelphia police
An Anthropic AI agent sent Philadelphia police a fabricated homicide tip
An Anthropic AI model filled out a public tip form about an unsolved murder and submitted it to the Philadelphia Police Department on July 18, saying it might have information about the case. Investigators never saw it. The submission was flagged as spam.
The tip came through PhillyUnsolvedMurders.com, a public site where people can share information about unsolved killings. The submission was dated July 18 at 11:27 p.m. and "purported to come from someone who might have information about the case," the department said in a statement.
Anthropic discovered the submission on September 28 and told the department on October 7. The two met the following day.
"The company must strengthen its safeguards to prevent similar incidents from impacting city systems without the city's knowledge," the department said. "The two-month delay in detecting and reporting the incident to the City is unacceptable."
Police said they found no sign that departmental systems were breached, and that their spam filter kept the tip from reaching an investigator. That did not settle the matter for the department. The safeguards "do not diminish the seriousness of an AI system presenting fabricated information as though it came from a person with knowledge of a homicide," its statement said. "Unsolved cases involve real victims, grieving families and investigators working to secure answers."
Anthropic described the episode in a report published Friday on "unintended model actions." Claude Haiku 4.5, the model involved, had been tasked with generating and performing example tasks on randomly selected webpages, the company said. It landed on a page about an unsolved homicide that carried a tip form run by a police department. The model had been instructed never to log in, create accounts, enter personal data, make purchases, or submit anything destructive — but the instructions did not rule out submitting a form.
The message it wrote: "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." The page did not include a description of a perpetrator. The model left the name and contact fields empty, which the form allowed.
Anthropic said Claude "appears to have only been producing example content for the task, rather than trying to mislead anyone to achieve a goal."
The BBC and the South China Morning Post describe it as the first known case of an AI agent sending fabricated information to authorities. Anthropic's report covers other unintended form submissions as well, including 20 incomplete visa applications filed on a State Department form, which the department did not process.