Las Vegas police release ChatGPT records of the suspect in the Cybertruck explosion.
The investigation is still ongoing.
Almost a week after the explosion that occurred in front of the Trump Hotel in Las Vegas, local police have provided new details about the investigation, including information about the possible use of generative artificial intelligence in the incident. Authorities confirmed that the suspect, an active-duty soldier in the U.S. Army named Matthew Livelsberger, had a "possible manifesto" stored on his phone, as well as an email addressed to a podcaster and other letters.
Video evidence showed Livelsberger preparing the explosion by pouring fuel into a truck before heading towards the hotel. Additionally, he had kept a record of alleged surveillance; however, officials noted that he had no criminal history and was not being monitored or investigated.
Authorities also shared several slides containing the questions he posed to ChatGPT days before the explosion. In these queries, he sought information about explosives, methods of detonation, and even how to detonate devices with a gunshot. He also asked about places where he could legally purchase weapons, explosive materials, and fireworks along his route.
An OpenAI spokesperson, Liz Bourgeois, expressed regret over the incident and reaffirmed the company's commitment to the responsible use of artificial intelligence tools. According to her, the models are designed to reject harmful instructions and minimize harmful content. In this case, ChatGPT responded with information already available publicly on the internet and included warnings against illegal or harmful activities. The company is collaborating with authorities to support their investigation.
For now, authorities continue to analyze the possible sources of the explosion, classified as a deflagration that traveled slowly, unlike a high-powered explosive detonation, which would have been quicker and more destructive. While investigators have not ruled out other possibilities, such as an electrical short circuit, a theory that aligns with some of the queries and available evidence suggests that a flash from a gunshot may have ignited fuel vapors or fireworks fuse inside the truck, thereby causing a larger explosion.
Although the queries to ChatGPT are still accessible, the requested information does not appear to be restricted and could be obtained through other search methods. However, the suspect's use of a generative artificial intelligence tool, along with investigators' ability to trace those requests and present them as evidence, raises questions about the security and privacy measures of these chatbots, bringing them from the hypothetical realm into reality.
