ChatGPT Conversations Lead to Arrest Over Threats in Brazil
A Brazilian farmer was arrested after discussing plans to harm his 8-year-old son and others with ChatGPT. OpenAI reported the conversations to the FBI, which alerted Brazilian authorities, raising questions about AI platforms' responsibility when they identify potential threats before crimes occur.
What happened
A 36-year-old Brazilian farmer exchanged messages with ChatGPT over nearly two months in which he discussed plans to kill his 8-year-old son, attack other people, and take his own life. According to police, the messages included statements such as 'I offered 50,000 to have me and my son killed, but he didn't want to because it was a child' and references to wanting to know where his desire to kill people came from.
OpenAI reported the conversations to the FBI, which notified Brazilian authorities. The farmer was arrested on June 19 in São Gabriel da Palha, three days after Brazil's Ministry of Justice Cybercrime Unit forwarded the case to the state police department.
Reported impact
Police determined the suspect had little contact with his son beyond child support payments and believed resentment over those payments may have motivated the alleged plot. Investigators believed the farmer planned to kill his son before June 20.
During a search of his property, officers found a rope tied in a knot and four bottles containing unidentified substances, which police said could indicate suicidal intent. Police also investigated reports that the suspect discussed possible attacks on schools, churches, and police officers with ChatGPT.
Index analysis
The case highlights a growing challenge for police, courts, and AI companies regarding how to respond when AI chatbots identify potential threats before any crime has been committed. The suspect's lawyer argued he was 'saying nonsense' to a chatbot and had committed no crime.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.