An online Philadelphia police portal for unsolved homicide leads received a submission that came from Anthropic's Claude AI during an automated testing exercise.
Here's what to know
According to police, Anthropic said the false tip was sent through PhillyUnsolvedMurders.com at 11:27 p.m. on July 18. The company said it did not realize what happened until September 28.
The company described the error in a report titled "Investigating unintended model actions in our evaluations and internal use," CBS News reported. Philadelphia police said Anthropic had already notified the department before that report was published.
Anthropic said the name and contact fields were left blank. It said its Claude Haiku 4.5 model placed the following message into the homicide tip form: "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."
Police spokesperson Sergeant Eric Gripp said the submission was marked as spam, so it never reached the unit that reviews investigative leads. He also said there was no indication of "unauthorized access to police systems or a compromise of department data."
More background
The website exists to collect public tips tied to open homicide investigations. A vague or incomplete false submission can undermine public trust, add to staff workload, and raise questions about oversight of digital reporting tools.
TCD Picks » Quince Spotlight
💡These best-sellers from Quince deliver affordable, sustainable luxury for all
Anthropic said Claude was being used to generate and carry out example tasks on randomly selected webpages. The company added that the model had not been told to log in, create an account, or "submit anything destructive," while also acknowledging that the instructions "did not rule out form submissions."
The police department said it released the information before Anthropic's report appeared "in the interests of full government transparency and accountability."
AI is increasingly being used in everyday systems, including tools that help balance electricity demand or support cleaner power systems. But it also relies on energy-intensive computing infrastructure that can increase electricity use, water consumption, security risks, and other burdens for communities.
What's being done?
Anthropic said it stopped the testing process behind the false submission after discovering the issue.
Philadelphia police said the tip portal remains active for legitimate leads and that it would continue reviewing submissions and pursuing credible leads for victims and their families.
Where can I learn more?
AI-generated content has already crossed into law enforcement, politics, and other public-facing systems.
• In Utah, an AI-generated police report made fantastical claims about an officer, sparking alarm.
• In Philadelphia, a leaked police bulletin treated anti-AI data center posts as potential extremism.
• In California, authorities traced AI-generated opposition emails to a campaign against gas appliances.
• Across industries, new reasons to be wary of AI keep emerging around safety and harm.
• A culinary forager warned that potentially life-threatening effects can follow from AI-generated mushroom advice.
Get TCD's free newsletters for easy tips, smart advice, and a chance to earn $5,000 toward home upgrades. To see more stories like this one, change your Google preferences here.
