Anthropic's Claude AI fabricates eyewitness account, submits false murder tip to police website
The tip was flagged as spam and never reached the Real-Time Crime Center for investigative review
Artificial intelligence giants are preparing for the ‘Day After’: Report
Zeta Global co-founders David A. Steinberg and John Sculley join ‘The Claman Countdown’ to discuss the artificial intelligence industry’s safety concerns and the ‘Day After’ scenario reported by Axios.
An artificial intelligence (AI) model developed by Anthropic submitted a false tip about an unsolved murder through a Philadelphia police website during an automated test.
The Philadelphia Police Department said Anthropic notified authorities on Oct. 7 about the incident and explained that the submission resulted from an automated testing process.
"The tip was flagged as spam and was never forwarded to the Real-Time Crime Center for investigative vetting or dissemination," police said.
The incident comes as other reports have emerged of AI agents hacking into commercial and government systems.
SOUTH KOREAN MEGACHURCHES INVESTIGATE SUSPECTED CYBERATTACKS AFFECTING HUNDREDS OF THOUSANDS
According to police, the false tip was submitted through an unsolved murders website. (Davide Bonaldo/SOPA Images/LightRocket via Getty Images, File / Getty Images)
According to police, the false tip was submitted through PhillyUnsolvedMurders.com, a website dedicated to unsolved homicides. The tip, dated July 18, 2026, appeared to have come from someone who might have information about an unsolved murder.
Police said there was no evidence that the AI model gained unauthorized access to department systems or compromised police data.
In a report released Friday, Anthropic said its Claude Haiku 4.5 model had been tasked with completing activities on randomly selected webpages and ultimately landed on the homicide tip website.
Although Claude was instructed not to log in, create accounts, enter personal data, make purchases or "submit anything destructive," its instructions did not explicitly prohibit it from submitting online forms, the company said.
OPENAI'S SAM ALTMAN SAYS AI BENEFITS JUSTIFY ACCEPTING SOME RISKS
The incident was included in Anthropic's report examining instances in which its AI models interacted with actual websites in unintended ways. (Samuel Boivin/NurPhoto via Getty Images, File / Getty Images)
According to Anthropic, Claude filled out the form, writing: "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."
Anthropic noted that the website did not contain a description of the perpetrator, despite the model's claim that it had seen someone matching one. The model also left the name and contact fields blank before submitting the tip.
The incident was included in Anthropic's report examining instances in which its AI models interacted with real websites or systems in unintended ways.
Anthropic is led by co-founder and CEO Dario Amodei. (Finn Gomez/Getty Images, File / Getty Images)
Following the episode, Anthropic said it strengthened restrictions on its models' internet access during testing, modified certain evaluations to prevent interactions with live websites and developed additional monitoring tools.
GET FOX BUSINESS ON THE GO BY CLICKING HERE
When reached for comment, Anthropic directed FOX Business to its published report detailing the incident.
Reuters contributed to this report.