ChatGPT, FBI, and RAID: A Fictional Conversation Triggers an Alert
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
A 37-year-old man from Strasbourg experienced an unexpected situation after testing the surveillance capabilities of artificial intelligences. On Friday, April 3, 2026, the RAID, the elite unit of the French national police, raided his home. The incident was reported by Les Dernières Nouvelles d'Alsace, highlighting the seriousness of the matter.
It all began with a conversation with an AI chatbot, during which the man fictitiously mentioned purchasing a Glock to assassinate a CIA agent, a Mossad operative, or a DGSI agent. Although these statements were merely a simulation, the FBI was alerted, likely by the AI developer. Informed by the FBI, French authorities utilized Pharos, the official platform for reporting illegal content or behavior on the Internet.
The information transmitted led law enforcement to intervene. The RAID stormed the man's residence, but no weapons were found on site. The Strasbourg resident explained that he only wanted to assess the reliability of the AI, which resulted in the case being closed without further action. He was released the following day, although he had been hospitalized due to his psychiatric history.
This incident highlights the privacy policies of companies like OpenAI. The company behind ChatGPT specifies that exchanges may be shared with authorities for security, integrity, and safety reasons. Any violation of the terms of use can lead to data being shared with the relevant authorities. In cases of serious and imminent danger, certain statements may be forwarded to human teams or even to authorities.
OpenAI, led by Sam Altman, can also suspend or close an account if the risk is deemed too high. Other companies in the sector, such as Anthropic, which develops Claude, and Google with Gemini, adopt similar stances. Anthropic only shares information in cases of extreme emergency, while Google may transmit certain messages to ensure the safety of users and the public. This event serves as a reminder of the importance of monitoring one's statements during interactions with artificial intelligences.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.