In Brazil, a 36-year-old farmer was arrested on suspicion of planning the murder of his eight-year-old son. It turned out that the man had been discussing his criminal intentions with the ChatGPT chatbot for more than two months, reports the website infohub.kz.
According to the investigation, the suspect had detailed conversations with the artificial intelligence, describing a plan to kill the child and then commit suicide. ChatGPT's safety system flagged these messages as potentially dangerous and forwarded the information to the FBI in the United States.
American law enforcement contacted Brazilian authorities, and the data was then passed to Ciberlab, the technology division of Brazil's Ministry of Justice, on June 16. Just three days later, police conducted an operation and arrested the suspect. According to the investigation, the man intended to carry out his plan on June 20 but was arrested four days before the planned date.
Investigators believe the motive may have been child support. According to police, the man wanted to avoid paying child support and had almost no relationship with his son. During a search of the suspect's apartment, four bottles with an unknown substance and a rope were found.
Initially, the man was taken into custody. However, on August 13, a court ordered his release because the investigation had dragged on. He is now required to wear an electronic bracelet and cannot approach his son or the boy's mother.
The man's lawyer, Pedro Pessy, stated that merely discussing such things with artificial intelligence should not automatically be considered evidence of a crime. "If talking to AI and spouting nonsense is a crime, then everyone should go to jail," he said.
Earlier, "Kursiv" wrote about a study that found ChatGPT, Gemini, Claude, and Grok massively cited erroneous sources.