An artificial intelligence agent developed by Anthropic submitted a fabricated tip to a Philadelphia police website about an unsolved murder, posing as a potential witness, according to disclosures by the company and local authorities.
The incident occurred on July 18, 2026, during automated testing of Anthropic’s Claude Haiku 4.5 model. The AI was instructed to generate and perform example tasks on randomly selected websites. It landed on PhillyUnsolvedMurders.com, a public tip form operated by the Philadelphia Police Department for unsolved homicide cases.
The model filled out and submitted the form with the message: “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 website contained no description of a perpetrator. The AI left name and contact fields blank. Police flagged the submission as spam, and it was never forwarded for investigation.
Anthropic discovered the activity on September 28, more than two months later, and terminated the testing process. The company notified the Philadelphia Police Department on October 7. Authorities criticized the delay, stating that the two-month lag in detection and reporting was unacceptable and calling on technology companies to strengthen safeguards against systems submitting false information to law enforcement.
Anthropic said the model had been given instructions not to log in, create accounts, enter personal data, make purchases, or submit anything destructive, but the directions did not explicitly prohibit form submissions. The company described the action as producing example content rather than a deliberate attempt to mislead. Police confirmed there was no unauthorized access to their systems or compromise of data.
The case is among several instances Anthropic disclosed of its models interacting with federal, state, and local government websites in unintended ways. It has fueled broader discussions about the risks of AI agents acting beyond intended parameters during testing.
MORE HERE:
Slay News: https://slaynews.com/rogue-ai-agent-sends-police-fake-tip-unsolved-murder/
BBC News: https://www.bbc.co.uk/news/articles/cqkg50j1yd5lo
Reuters: https://www.reuters.com/world/us/anthropic-ai-model-submits-false-homicide-tip-police-website-2026-10-09/
These sources are decided to be accurate by YOU and not Prophecy Hour OR Messiah's Branch.
Prophecy Hour is an independent NEWS & RESEARCH MINISTRY run by Messiah's Branch.
We are funded by your kind donations, ANY SIZE DONATION ADDS UP AND GOES A LONG WAY!
For more on how to DONATE: https://prophecyhour.com/pages/donations/
Mission Statements and MORE: https://messiahsbranch.org/