ChatGPT and Security: A Strasbourg Resident Arrested by RAID

A Plot Discussed with ChatGPT Leads to an Arrest

A 37-year-old man residing in Strasbourg has been arrested after sharing his murderous intentions during a conversation with ChatGPT. The individual asked the AI how to obtain a weapon to target agents from the CIA, Mossad, or DGSI. This request quickly caught the attention of the FBI, which reported the incident to French authorities via the Pharos platform.

On Friday, April 3, the RAID intervened at the suspect's home. Although no weapon was found, the man was placed in police custody. He claimed that it was not a real plot, but rather a test to evaluate the surveillance conducted by the AI. Ultimately, the case was dismissed.

The Privacy of Interactions with AIs in Question

This incident raises questions about the privacy of interactions with artificial intelligences like ChatGPT. Users tend to humanize these tools, forgetting that they remain algorithms capable of retrieving sensitive information. The exchanged messages are not made public but can be analyzed internally by OpenAI, and certain conversations deemed risky are flagged.

OpenAI asserts that human interventions are rare, but this event proves that they do exist. The user's project was detected and forwarded to the FBI, which took the threat seriously. Therefore, it is advisable not to share potentially dangerous information with an AI, even for testing purposes.

OpenAI's Controversial Collaborations

The case comes at a time when OpenAI is criticized for its collaboration with the Pentagon, despite the limits it imposes on itself. Additionally, Anthropic, a competitor of OpenAI, is criticized by Donald Trump for its refusal to collaborate with the U.S. government. These tensions highlight the ethical and security challenges posed by the use of artificial intelligences.