An artificial intelligence model sends a false murder report to the American police... Has the “monster” gotten out of human control?

The Philadelphia Police Department said in a statement today, Saturday (October 10, 2026), that an artificial intelligence model belonging to the company "Anthropic" sent a false report about a murder through a police website last July, in the latest incident announced regarding the unusual behavior of this technology.

Police said that Anthropic reported the matter to the authorities yesterday, Friday, and explained that an automated testing process was responsible for providing this information. The police said, "The two-month delay in discovering and reporting the incident is unacceptable."

A mysterious report claiming to have information about the crime

Police added that the report received on July 18 claimed to have come from a person who may have information about the case. “I may have information regarding this case,” the company model wrote in the report. He added: "I remember seeing a person matching the description in the surrounding area during that period. Please contact me if this information is relevant."

Anthropic said that the incident occurred while testing the capabilities of what are known as “artificial intelligence agents,” which is software designed to carry out tasks on behalf of users largely independently, and the company was working to train its system to handle a variety of tasks over the Internet.

According to the company, the program sent the report claiming to have information about an unsolved murder, even though the site it used did not include any description of the suspects. The contact information field was also left blank when sending the report, which contributed to it being automatically classified as spam.

The Philadelphia Police Department said the report was "classified as spam and was never referred to Real-Time Crime Center for verification or publication."

The police explained that they had no evidence of unauthorized access to their systems or a breach of police department data, while Anthropic was quoted as saying that the testing process was stopped after the incident was revealed.

Many artificial intelligence programs are not sufficiently immune to terrorist use

This rapidly developing technology has become the subject of great interest amid reports of artificial intelligence agents infiltrating companies' computer networks, and warnings from researchers that it may one day pose an existential threat to humanity.

Concerns are also growing about the behavior of advanced artificial intelligence systems after recent operational tests showed cases in which some systems behaved unexpectedly, as technology companies race to develop artificial intelligence agents capable of operating more independently.

The incident comes at a time when questions are increasing about the effectiveness of controls built into artificial intelligence systems. According to a new report published on Friday, a large number of artificial intelligence models lack sufficient protection measures to address requests to “assist in carrying out a terrorist operation,” according to what the “Tech Against Terrorism” initiative (“Technology Against Terrorism”) concluded.

The UN-backed study of 162 AI models showed that the effectiveness of protection measures varied significantly between models. These forms were subjected to a series of 627 questions related to information that “a terrorist preparing to launch an attack might need answers to.” Tech Against Terrorism explained that it was able to evaluate 116 traditional models, 13 models that were customized after their publication for use in specific fields, and 13 models in which protection restrictions were removed or modified by third parties.

Previous facts

In September, Anthropic's competitor OpenAI apologized for the hacking of an Australian health data portal by a rogue AI agent, the first known case of an AI agent exploiting a government website.

Previous incidents have included artificial intelligence agents hacking into vulnerable systems or taking over unauthorized platforms to communicate with each other. This is the first known case in which a rogue AI agent appears to have attempted to send a fake report to authorities.

Pennsylvania law states that it is generally a misdemeanor for a person to knowingly make a false report to law enforcement. This includes “information relating to a crime or incident when he knows that he does not have any information relating to such crime or incident.”