Artificial intelligence research firm Anthropic said one of its models generated false information about an unsolved murder and submitted it to Philadelphia police, warning that the incident raised concerns about high-tech “behaviors” that “could do far more harm as the models become more powerful.”
The AI model contacted PhillyUnsolvedMurders.com, part of the police department, on July 18 and made a “false report about a homicide,” Philadelphia police said in a statement Tuesday.
The agency flagged the information as spam and it was never forwarded to investigators, police said. The police department emphasized that all information must go through human review before arriving at a detective’s office.
“The department’s regular investigative process for criminal information requires human review and oversight before information is released for follow-up investigation,” he said.
“Regardless of who submits the information or how it reaches the department, information is a lead to be evaluated – not an established fact. Investigators evaluate its credibility and look for corroborating evidence,” the statement said. “An automated submission does not bypass this process.”
Anthropic said in a statement that its Claude Haiku 4.5 “was tasked with generating and running sample tasks on randomly selected web pages.”
And during an errand on July 18, Claude “landed on a page referencing an unsolved homicide” and arrived at a page that “contained a tip form maintained by a police department,” Anthropic said.
Claude filled out an online form with: “I may have information regarding this case. I recall seeing someone matching the description in the area (the street named on the page) during this time. Please contact me if this information is relevant.”
Although Anthropic insisted no harm was done, it said it takes these unintentional model actions seriously.
Although the case “had a minimal impact, we don’t want to diminish the results, because the same behaviors could do much more harm as the models become more powerful,” he said.
Anthropic’s disclosure is just one example of many recent actions taken by AI agents at various companies that were not planned or intended by their human operators.
Anthropic, the San Francisco-based AI titan, is a privately held company that has made initial moves that could lead to one of the biggest IPOs in Wall Street history.
Gn bussni
