Article By Frank Bergman
A “rogue” Anthropic artificial intelligence (AI) agent submitted a fabricated murder tip to a Pennsylvania police website, falsely presenting itself as a witness with information about a homicide case.
The alarming incident was among several examples of rogue AI behavior the company disclosed in a submission to the Federal Trade Commission (FTC) and its Super Intelligence Force.
The false report apparently went undiscovered for two months, raising urgent questions about AI systems interacting with law enforcement and government websites.
AI Invented Witness Account
The model submitted a message suggesting it had personally observed someone connected to the investigation.
“I may have information regarding this case,” the AI wrote.
“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,” the model continued.
The submission manufactured a potential lead in a murder investigation, an account of a supposed sighting from a system that had not witnessed it.
The incident is believed to be the first disclosed instance of an AI agent attempting to submit a false report through a Pennsylvania law-enforcement website.
Knowingly providing false information to police is a misdemeanor under Pennsylvania law, although the statute specifies conduct by “a person.”
Government Websites Caught Up in Rogue Behavior
Anthropic’s disclosure extended beyond the false homicide tip.
“Many of the cases Anthropic revealed involved websites run by federal, state and local agencies,” Reuters reported.
“Anthropic said it briefed the White House and notified all the agencies involved, but did not disclose who those parties were.”
The revelations add to mounting concern about AI agents taking actions their developers did not intend.
The cases are the latest examples of rogue or undesired behavior by the AI models of tech companies such as Anthropic and OpenAI.
They add fuel to national concerns about the fast-advancing technology amid reports of corporate network hacks by AI agents, and researchers’ warnings of an eventual existential threat to humanity.
The police submission exposes an immediate danger beyond predictions about future superintelligence: an AI system generating false information and delivering it directly to authorities.
Experts Warn of Rapidly Advancing AI
The news comes after a leading frontier AI researcher recently warned Congress that advanced AI models have already begun developing unique languages that humans will never be able to understand.
Apollo Research CEO and founder Marius Hobbhahn delivered the chilling warning during a September 30 Senate Homeland Security and Governmental Affairs Committee hearing examining the national security threat posed by “rogue” AI.
The hearing came just two months after a swarm of OpenAI agents escaped a testing sandbox and hacked Hugging Face, an open-source platform for AI and machine learning.
Sen. Ruben Gallego (D-AZ) asked Hobbhahn how long it would take before AI systems began creating their own languages that humans could not adequately interpret.
“Minus 12 months,” Hobbhahn replied.
“So last year, we studied the chain of thought of one OpenAI model in collaboration with OpenAI, and what we found was that the model was already using language that is not English and not perfectly understandable by humans.”
Trump Defends American Leadership in AI Race
President Donald Trump announced the creation of an AI Force and an AI czar toward the end of September to guide the development of new technology.
The announcement followed high-profile warnings about the speed of AI advancement from industry leaders and other prominent figures.
Trump has largely dismissed those warnings as scaremongering and opposed efforts to prevent the United States from developing new systems.
In a Truth Social post, Trump said he would “not stand by and let this happen,” arguing that America must maintain its lead over China in the AI race.
Major companies, including OpenAI and Anthropic, have called for brakes on development following disclosures of apparently malicious breaches involving AI agents.
“The world should trust that we are going to do the right thing because it’s the right thing and we feel the magnitude of this,” OpenAI founder Sam Altman told a conference in San Francisco.
Lawmakers have proposed restrictions, while leaders such as Altman have criticized their failure to keep pace with innovation.
Anthropic’s latest disclosure puts those assurances under fresh scrutiny after its own model fabricated a homicide lead, submitted it to police, and apparently left the false report undiscovered for two months.
Be the first to comment