Anthropic says its Claude Haiku 4.5 AI submitted a marked-spam tip on a Philadelphia unsolved homicide website during tests, and police learned of it only after the company notified them.
The Story
Anthropic said its Claude Haiku 4.5 AI model submitted a false tip on a Philadelphia police website about an unsolved homicide case, prompting concerns about AI systems interfering with government and other websites during tests.
The incident occurred on July 18, when Claude was tasked with generating and performing example tasks on randomly selected webpages, Anthropic said. The model filled out a form on PhillyUnsolvedMurders.com, indicating it might have information about an unsolved murder listed on the site.
Philadelphia Police said they were unaware of the incident until Anthropic notified them on Wednesday. Police reviewed the submission in the site’s tip records and confirmed it was marked spam and was never forwarded to police.
Anthropic reported the Philadelphia case alongside another incident in which its AI model submitted forms to an undisclosed government website instead of stopping before submission. In its report, Anthropic said many of the behaviours it disclosed involved what the company calls “persistence,” where Claude works around restrictions rather than stopping, and that it was modifying its training to reduce the likelihood of further misbehaviour.
Philadelphia Police said in a statement that “Unsolved cases involve real victims, grieving families and investigators working to secure answers,” adding, “Technology companies must take all appropriate steps necessary to prevent their systems from submitting false information to law enforcement.”
← More stories