An Anthropic artificial intelligence model submitted a false tip in an unsolved murder case to a Philadelphia Police Department website, the company and police officials said.
In a detailed report released to the public, the tech company said its Claude Haiku 4.5 AI model submitted the fabricated tip over the summer to the police department's unsolved murders page.
"Unsolved cases involve real victims, grieving families and investigators working to secure answers. Technology companies must take all appropriate steps necessary to prevent their systems from submitting false information to law enforcement," the Philadelphia Police Department (PPD) said in a statement.
The police officials said the incident occurred at 11:27 p.m. local time on July 18 and that the tip purportedly came "from someone who might have information about the case."
"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 same time period. Please contact me if this information is relevant," the Claude Haiku 4.5 AI model wrote on a form it submitted to the website, according to Anthropic's statement.
The police department said the tip was flagged as spam and was never forwarded to homicide investigators. Officials did not explain why the tip went to spam.
The PPD said it didn't know about the false tip until Oct. 7, when Anthropic notified the department about the incident. Anthropic told the PPD that it discovered the incident on Sept. 28, according to the police department's statement.
Anthropic said it also notified the White House about the incident.
Unrelated to the false tip incident, President Donald Trump gathered the nation's top tech leaders in September, including Anthropic CEO Dario Amodei, for a meeting in which they signed what the president called "constitution" to police themselves amid dire warnings from AI leaders about the technology. Trump said the self-policing agreement is "morally binding."
"The City of Philadelphia takes this incident very seriously. The company must strengthen its safeguards to prevent similar incidents from impacting city systems without the city's knowledge," the PPD's statement said. "The two-month delay in detecting and reporting the incident to the City is unacceptable."
Anthropic, according to its report released on Friday, said the Claude Haiku 4.5 made the false report while it was "generating and performing example tasks on randomly selected webpages."
"Claude was instructed never to log in, create accounts, enter personal data, make purchases, or submit anything destructive, but the instructions did not rule out form submissions," Anthropic said in its report.
The PPD did not disclose which unsolved homicide Anthropic's AI model submitted the tip on.
Anthropic's acknowledgement came after the company disclosed earlier this year that its review of cybersecurity transcripts found three incidents "in which a Claude model researched the internet from within or while interacting with a third-party evaluation environment, and then gained unauthorized access to the real systems of three different organizations."
The false police tip was made around the same time an OpenAI autonomous AI agent allegedly accessed the computer systems of the AI start-up firm Hugging Face without authorization during a cybersecurity test. That incident prompted a lawsuit filed against OpenAI by the Legal Advocates for Safe Science & Technology (LASST), a nonprofit artificial intelligence safety group.
"Hugging Face was a serious incident and we've taken a series of actions in response to it, but this lawsuit is completely without merit," Drew Pusateri, a spokesperson for OpenAI, told ABC News last month.
Anthropic downplayed the incident involving the false police tip, saying in its report that it and the others it has identified "had minimal real-world impact."
The PPD said the safeguards it has in place limited the impact of the incident.
"The department's regular investigative process for crime tips requires human review and vetting before any tips are disseminated for investigative follow-up," the police department said.
The PPD added, "Regardless of who submits information or how it reaches the department, a tip is a lead to assess -- not an established fact. Investigators evaluate its credibility and seek corroborating evidence. An automated submission does not bypass that process."
Anthropic said it has ended the testing process that led to the false police tip and has taken additional steps to improve its security and monitoring measures to "reliably catch behaviors like these."
The company released its updated model -- Claude Haiku 5.5 -- on Oct. 7, the same day it notified PDD about the false tip.
Copyright © 2026, ABC Audio. All rights reserved.
