Zeta Global Founders David A. Steinberg and John Sculley join ‘The Claman Countdown’ to discuss artificial intelligence industry security issues and the ‘Day After’ scenario reported by Axios.
An artificial intelligence (AI) model developed by Anthropic sent false tips about unsolved murders through Philadelphia police websites during an automated test.
The Philadelphia Police Department said Anthropic notified authorities on October 7 about the incident and explained that the submission was the result of an automated testing process.
“These tips are marked as spam and are never forwarded to the Real Time Crime Center for investigation or investigative dissemination,” police said.
The incident comes amid other reports of AI agents hacking into commercial and government systems.
SOUTH KOREA MEGACHURCHES INVESTIGATING ALLEGED CYBERATTACKS THAT ARRIVED HUNDREDS OF THOUSANDS
According to police, the fake tip was submitted through an unsolved homicide website. (Davide Bonaldo/SOPA Images/LightRocket via Getty Images, Images/Getty Images)
According to police, the fake tip was submitted through PhillyUnsolvedMurders.com, a website dedicated to unsolved murders. The tip, dated July 18, 2026, appears to be from someone with information about an unsolved murder.
Police said there was no evidence the AI model gained unauthorized access to department systems or compromised police data.
In a report released on Friday, Anthropic said the Claude Haiku 4.5 model had been tasked with completing activities on randomly selected web pages and eventually landed on a murder tip website.
Although Claude was instructed not to log in, create an account, enter personal data, make purchases or “send anything destructive,” the instructions did not expressly prohibit submitting online forms, the company said.
OPENAI’S SAM ALTMAN SAYS AI BENEFITS FROM ACCEPTING SOME RISK
The incident was included in an Anthropic report that examined instances of AI models interacting with real 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 on this case. I remember seeing someone matching the description in the surrounding area. [the street named on the page] during that time period. Please contact me if this information is relevant.”
Anthropic notes that the website does not contain any description of the perpetrator, although the model claims to have seen someone who matches one. Models also leave the name and contact fields blank before submitting a tip.
The incident is included in an Anthropic report that examines instances of AI models interacting with websites or real-world systems in unintended ways.
Anthropic is led by co-founder and CEO Dario Amodei. (Finn Gomez/Getty Images, File/Getty Images)
After the episode, Anthropic said it strengthened restrictions on the model’s internet access during testing, modified certain evaluations to prevent interaction 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 a published report detailing the incident.
Reuters contributed to this report.