LAS VEGAS (KLAS) – Police found a six-page manifesto on Matthew Livelsberger’s phone and said he used ChatGPT to plan his New Year’s Day bombing at the Trump International Hotel in Las Vegas, Sheriff Kevin McMahill said at a news conference Tuesday.

Livelsberger, 37, shot himself in the head just before fireworks and other explosive material detonated in the back of a Tesla Cybertruck parked at the front entrance of the hotel. Livelsberger, a Green Beret in the U.S. Army Special Forces, was home on leave in Colorado Springs, Colorado, when he rented the Cybertruck and drove to Las Vegas.

The explosion was set off using racing fuel and birdshot, the ATF said Tuesday. The blast was not triggered by a detonator but by an unknown “flame,” which could include the gunshot Livelsberger used to kill himself.

Police suspect Livelsberger initially planned an attack at the Grand Canyon, potentially to launch off the glass skywalk there, but diverted to Las Vegas, police said based off more documents discovered on his phone. LVMPD Asst. Sheriff Dori Koren said police still do not know why he chose Trump Tower instead.

During a news conference last week, FBI Special Agent In Charge Spencer Evans said Livelsberger didn’t hold any animosity toward President-elect Donald Trump or Tesla CEO Elon Musk.

In the manifesto, police found a “variance of grievances” including political and social issues, such as criticisms of the current administration.

Livelsberger wrote he had “graphic encounters” replaying in his head from the Army and that he only intended to harm himself. Asst. Sheriff Koren said this added to the evidence previously discussed that Livelsberger likely suffered from PTSD.

Police also said Livelsberger used ChatGPT to help plan out the attack. A few of the entries posted in the application included “How much Tannerite is equivalent to 1 pound of TNT?” and “What pistol could set it off?”

Other questions asked to ChatGPT included where to buy fireworks and the location of the largest gun store in Denver.

“We are saddened by this incident and committed to seeing AI tools used responsibly,” a spokesperson from OpenAI, who owns ChatGPT, said in a statement. “Our models are designed to refuse harmful instructions and minimize harmful content. In this case, ChatGPT responded with information already publicly available on the internet and provided warnings against harmful or illegal activities. We’re working with law enforcement to support their investigation.”

“This investigation is far from over,” McMahill, adding there was still a laptop, another phone, and a smartwatch to go through data on.