AI saves a child's life
Identified by the company's security system, the attempted homicide was reported to the police
VIGILANCE ChatGPT: the security system noticed the plot and alerted the FBI (./Getty Images)
Amid the intense debate about the risks of artificial intelligence (AI), a case in Espírito Santo shows that the technology can also be a safety tool. What would have been a private conversation between an individual and a chatbot turned into the key piece to prevent a murder. Thanks to the monitoring mechanisms used by OpenAI, the company responsible for ChatGPT, a plan to kill a child in Brazil was identified and communicated to American authorities, who contacted the Brazilian police, who managed to act and prevent a filicide.
The case came to light this Thursday, the 25th, when the Civil Police of Espírito Santo revealed the arrest of a 36-year-old man, accused of planning to kill his own son, an 8-year-old child, to avoid paying child support to his ex-partner. According to investigations, he used ChatGPT as a kind of diary, recording his intentions, doubts and the planning of the crime. The man intended to commit the crime on the 20th.
The messages caught the attention of the platform's security systems, which triggered the FBI, the federal police of the United States. The information was shared with Brazilian authorities, allowing the Specialized Cybercrime Repression Police Station (DRCC) to locate and arrest the suspect in the rural area of São Gabriel da Palha, one day before the date on which the murder was to be executed.
The man confessed to the plan during interrogation. The episode illustrates a little-known aspect of generative artificial intelligence. Although conversations with chatbots are largely private, companies like OpenAI maintain systems designed to identify content that indicates imminent risk to life, terrorism, child exploitation or other serious threats. In such cases, internal protocols allow information to be forwarded to the competent authorities, always within legal requirements and when there is a concrete risk to people's safety.
The case reinforces a movement that is gaining strength among large technology companies. Due to social pressure and many lawsuits, digital platforms have begun to invest heavily in security teams and in artificial intelligence models capable of detecting patterns associated with extreme violence, suicide, child sexual exploitation and attack planning. The goal is to transform the enormous processing capacity of these systems into a preventive tool, capable of identifying critical situations before they turn into tragedies. They also ensure that it is not seen as a lawless territory where anything is allowed.
