The suspect in the New Year’s Eve bombing at the Trump International Hotel in Las Vegas used a popular chatbot called ChatGPT to plan the attack, police say, in what is the first known case of the chatbot being involved in making explosives.
The suspect in the New Year’s Eve bombing at the Trump International Hotel in Las Vegas used a popular chatbot called ChatGPT to plan the attack, police say, in what is the first known case of the chatbot being involved in making explosives.
According to Reuters, the suspect used ChatGPT to find out how much explosives were needed to carry out the explosion.
Last week, authorities identified the man found dead inside a Tesla Cybertruck as Matthew Livelsberger, a 37-year-old U.S. Army soldier from Colorado Springs. The FBI says he acted alone, and the incident appeared to be a suicide.
Las Vegas police previously said the Cybertruck explosion was the first incident in the United States where ChatGPT was used to create an explosive device. Skeptics of artificial intelligence warn that it could be used for malicious purposes, and the Las Vegas attack could be another reason for criticism.
«It is particularly noteworthy that we now have clear evidence that the suspect used ChatGPT’s artificial intelligence to plan his attack,» Sheriff Kevin McMahill of the Las Vegas Police Department said at a press conference.
ChatGPT developer OpenAI said the company is «committed to the responsible use of AI tools» and that its «models are designed to reject harmful instructions.»
«In this case, ChatGPT responded with information that was already publicly available online and provided a warning against harmful or illegal actions,» the company said in a statement.
The FBI also noted that the suspect had no animosity toward US President-elect Donald Trump and likely suffered from post-traumatic stress disorder.
Recall that OpenAI previously announced a new family of artificial intelligence models, o3, which the company claims are more advanced than o1 or any others released before it. OpenAI claimed that they have become more in line with the company’s security principles. In particular, the speed at which o1 answered «dangerous» questions — at least those that OpenAI considered dangerous — was reduced, while improving its ability to answer «normal» ones.
