Before Las Vegas, Intel analysts warned that bomb makers were turning to artificial intelligence

Six days before he committed suicide outside the main entrance of the Trump International Hotel in Las Vegas, highly decorated US Army Green Beret Matthew Livelsberger of Colorado consulted an artificial intelligence using a series of prompts. He rented a Cybertruck and turned it into a four-ton vehicle carrying explosives. According to documents obtained exclusively by WIRED, US intelligence analysts have been warning about this exact scenario for the past year, and among their concerns is that AI tools could be used by racially or ideologically motivated extremists to target critical infrastructure, particularly power structures. network.

“We knew that artificial intelligence was going to change the game, really, all of our lives at one point or another,” Las Vegas Metropolitan Police Department Sheriff Kevin McMahill told reporters Tuesday. “Of course, this is an anxious moment for us.”

Transcripts of OpenAI’s exchanges with ChatGPT show that Livelsberger, 37, was on his way to Las Vegas to collect as much explosive material as he could legally, as well as learn how to detonate it using a Desert Eagle weapon found there. Cybertruck after his death. Screenshots shared by McMahill’s office show Livelsberger asking ChatGPT for information about Tannerite, a reactive compound typically used for target application. In one such survey, Livelsberger asks, “How much tannerite is equal to 1 pound of TNT?” He follows up by asking how he can ignite at “point blank”.

Documents obtained by WIRED show that concerns about the threat of artificial intelligence being used to help commit serious crimes, including terrorism, are widespread among US law enforcement agencies. They report that the Department of Homeland Security continues to warn about domestic extremists who rely on the technology to “create bomb-making instructions” and develop “common tactics for carrying out attacks against the United States.”

Violent extremists are increasingly turning to tools like ChatGPT to mount attacks aimed at bringing down American society through acts of domestic terrorism, the records, which are not classified but are limited to government officials.

according to notes According to what investigators found on his phone, Livelsberger intended the blast as a “wake-up call” to Americans urging them to reject diversity, embrace masculinity and rally around President-elect Donald Trump, Elon Musk and Robert F. Kennedy Jr. He also called on Americans to “tough reconstruction,” calling for Democrats to be purged from the federal government and the military.

While McMahill claimed that Tuesday’s incident in Las Vegas may be the first time that “ChatGPT has been used within the United States to help individuals set up a specific device,” federal intelligence analysts say white supremacists and extremists associated with the Acceleration movement are now rapidly gaining access to the Internet. – they share quick access. hacked versions of AI chatbots to create bombs to launch attacks against law enforcement agencies, government facilities and critical infrastructure.

In particular, the memos highlight the vulnerability of the US power grid, a popular target among extremists.Terrorgram,” a loose network of encrypted chat rooms that hosted a series of violent, racially motivated individuals seeking to destroy American democratic institutions. Documents shared only with WIRED were originally obtained by Property of the Peoplea non-profit organization focused on national security and government transparency.