A 36-year-old farmer from Brazil was taken into custody after spending nearly two months discussing plans to kill his eight-year-old son with ChatGPT. Authorities suspect that one of his motivations may have been a desire to avoid ongoing child support payments. Throughout his interactions with the AI, the man inquired about the effects of toxic substances. OpenAI's safety protocols identified these discussions as a potential threat, leading to a notification to the FBI, which subsequently relayed the information to Brazilian law enforcement. OpenAI has stated that it may inform authorities when conversations indicate an "imminent and credible risk of harm to others." The man was apprehended on June 19, with police asserting that he planned to harm his son by the following day. During a search of his residence, law enforcement discovered four containers containing an unknown substance along with a rope. After spending 54 days in custody, a court mandated his release due to the extended duration of the investigation without conclusive results. He is now required to wear an electronic ankle bracelet and is prohibited from contacting both his son and the child's mother. The investigation is still active, and so far, no charges have been filed against him. His attorney, Pedro Paulo Pessi, remarked, "If conversing with AI and saying absurd things is a crime, then everyone could be arrested." The situation also raises critical questions regarding the evidence collected. While police obtained the messages sent by the man to ChatGPT, they did not receive the AI's replies. Investigators and legal analysts have noted that the absence of these responses complicates the evaluation of the conversations in their entirety. The debate continues on whether AI companies should be obligated to report such interactions to law enforcement.
Informational material. 18+.
