OpenAI alerts Federal Police to death threats against Lula on ChatGPT

OpenAI notified Brazil's Federal Police that at least two ChatGPT users expressed intent to attack President Luiz Inácio Lula da Silva. The company classified the conversations as a possible emergency involving risk of death or serious injury and shared the dialogue logs with investigators in early October.

One user had messaged the chatbot since January and reported seeing Lula on a campaign stage roughly 400 metres away in September. The individual then wrote about wanting to shoot the president in the head and asked what legal penalties would follow, also sending a photo of a tower offering a clear line of sight.

“possible emergency with risk of death or serious physical injury”

OpenAI closed at least one of the accounts and alerted the Superior Electoral Court under its electoral compliance plan. The Federal Police forwarded the material to the Ministry of Justice, which must decide whether to open two inquiries — one for threats and insult, another specifically for threatening the president.

The Ministry of Justice and Federal Police have not commented on the case. OpenAI says it issues preventive alerts whenever conversations indicate real and imminent risk of harm to others.