Palisades fire defendant’s AI query sparks arson doubts

In federal court, an ATF agent testified that arson defendant Jonathan Rinderknecht asked ChatGPT whether “a fire lit because of your cigarettes” could make him responsible, while prosecutors allege he started the Lachman fire before it later fed the Palisades
Minutes after a fire erupted above Pacific Palisades on New Year’s Day 2025, Jonathan Rinderknecht asked ChatGPT a question that would later become part of a courtroom portrait of his mind.
“Are you at fault if a fire is lit because of your cigarettes?” the Uber driver wrote, according to evidence described Thursday in federal court as the government prepared witnesses in his arson trial.
That moment—quiet. digital. and contained to a single screen—has taken on a different weight as prosecutors lay out a theory that the Lachman fire was maliciously set just after midnight and then burned underground for days before reigniting on Jan. 7 as the Palisades fire, which became the most destructive wildfire in Los Angeles history.
The attorney for Rinderknecht. Steve Haney. has framed the case in a different direction. arguing that the government’s story leans heavily on motive and mental state while offering no proof that Rinderknecht intentionally started any blaze. Federal prosecutors have not yet presented evidence that he intentionally burned anything. and the evidence described Thursday featured the sharp contrast between what the government says was an act and what defense counsel says the records do not show.
Prosecutors say the Lachman fire erupted around midnight on New Year’s Day 2025 near Skull Rock in Temescal Canyon and burned underground for days. before it reignited Jan. 7 as the Palisades fire. The fire killed 12 people. destroyed more than 6. 500 structures across the Palisades and Malibu. and caused billions of dollars in damage. authorities allege.
In the courtroom, U.S. Alcohol, Tobacco, Firearms and Explosives agent Michael Montevidoni testified over two days on the stand about Rinderknecht’s movements and mental state in the moments investigators say he started the Lachman fire and the day after.
Montevidoni described texts, screen grabs, recordings, and other records recovered from a search of Rinderknecht’s cellphones, depicting a man who could be lonely and angry—angry at billionaires and ex-romantic partners—while also appearing to worry about his own declining mental health.
Prosecutors allege that Rinderknecht “maliciously” started the Lachman fire, pointing to statements he made to investigators about resentment toward the wealthy and to a lengthy conversation log with ChatGPT that prosecutors said reflected frustration with his financial situation and social life.
The government also pointed to the recovery of a green BIC barbecue lighter in Rinderknecht’s car that appeared in a video the defendant filmed near the fire scene on the night the Lachman blaze erupted.
But Haney’s approach has been to press a different gap: the absence of steps that would show arson planning.
“You never saw any prompts or any searches on any of his devices … on how to plan to burn something,” Haney said Thursday morning, challenging the evidence Montevidoni was presenting.
Haney pressed that point later in the morning. asking the ATF agent directly: “You never found any searches in any of the devices regarding arson. right?” He asked whether investigators found any internet searches for deliberately starting a fire—whether information about an incendiary device. lighter. torch. or flare—or whether there were purchases of fire-starting materials.
Montevidoni responded “no” each time.
In testimony, Montevidoni said he had not seen any prompts or searches showing how to plot a fire, even as he described hours spent detailing Rinderknecht’s conversations with ChatGPT. Montevidoni said Rinderknecht’s discussions with the chatbot were akin to a journal.
In July 2024. Montevidoni testified. Rinderknecht demanded the chatbot generate an image depicting wealthy elites dining extravagantly on one side of a wall while the world burned beyond the barricade. Rinderknecht prompted ChatGPT to create more than 40 images similar to that. and he cursed out the bot when it did not produce what he wanted or when it noted he was violating OpenAI’s content creation policies. according to images displayed in court.
Montevidoni also testified that Rinderknecht wrote to ChatGPT about a belief that the wealthy are “enslaving” society and causing harm to the planet. In messages shown in court, Rinderknecht wrote that he was struggling to accept that he could not do anything about it.
“Why am I so angry all the time,” he asked in one message.
“The cost of life out here brings me anxiety. My loneliness brings me anxiety,” he wrote in another. “I’m very lonely. Like, very, very lonely.”
The record described by Montevidoni included internet searches and web history tied to Luigi Mangione, the suspect in the killing of UnitedHealthcare’s chief executive. Montevidoni testified that Rinderknecht searched “Free Luigi” and “reddit lets kill all billionaires” in December 2024.
Montevidoni later testified that Rinderknecht also used ChatGPT to look for information about the home address and security protocols of DoorDash CEO Tony Xu.
The court testimony also traced a more personal thread into the hours leading up to the fire.
On Dec. 30. two days before the fire. Montevidoni said Rinderknecht reached out to a woman he had previously been in a romantic relationship with. asking her to spend New Year’s Eve with him. He was rebuffed. and then—according to Montevidoni’s testimony—Rinderknecht began sending vile messages to the woman from another phone. The messages included comments about her weight, calling her a “pathetic pig” and telling her “you got no worth.”.
Minutes before midnight, Montevidoni testified, Rinderknecht called an ex-boyfriend who he lived with in the Palisades from 2020 to 2022. Phone records showed the ex-boyfriend ignored the call. Montevidoni testified that Rinderknecht called again after the fire erupted.
At 12:12 a.m., he made two quick phone calls to 911, among many calls that L.A. County authorities received reporting the Lachman fire. Montevidoni said both calls lasted only seconds. He described Rinderknecht as panicked but brief, and said he did not say anything about how the fire started.
Two minutes later, Montevidoni said, Rinderknecht asked ChatGPT if someone would be responsible for the fire started by discarded cigarettes.
Montevidoni testified that Rinderknecht “screen-recorded” both the 911 calls and his ChatGPT prompt, describing it as odd.
“To us, it seems like he was trying to create an alternative reason for why the fire was started and have a record of that to show at a later period,” Montevidoni told the court.
Afterward, subsequent cellphone videos and surveillance cameras showed Rinderknecht driving around the neighborhood near the fire for almost two hours, sometimes trailing L.A. Fire Department vehicles.
Montevidoni testified that the next morning, Rinderknecht’s Google search history showed him repeatedly searching for information about the Lachman fire.
When he returned home to his North Hollywood apartment around 3:45 a.m., records from Rinderknecht’s phone showed he played a French hip-hop song he had become obsessed with. Montevidoni said the song’s video includes money set on fire.
Prosecutors have described a timeline that stretches across the holiday and into catastrophe: the Lachman fire erupted around midnight on New Year’s Day 2025. burned underground for days. and reignited on Jan. 7 as the Palisades fire. In the court’s account of Rinderknecht’s behavior. prosecutors are working to connect an alleged start to digital activity that they say suggests intent.
Defense counsel is pushing back just as hard, insisting the evidence he has heard does not show the planning that would make arson proof.
“It is unclear if Rinderknecht will testify at trial,” the proceedings have so far reflected, with Haney telling the court Thursday that the prosecution’s entire case appears focused on his client’s despair rather than the cause of the fire or proof that Rinderknecht was the person who set it.
Haney said the government has not shown searches on his devices about arson or deliberate fire-starting, including any evidence of purchases of fire-starting materials.
If convicted as charged, Rinderknecht faces 45 years in federal prison.
The courtroom testimony also described what investigators encountered as authorities moved in quickly after the Palisades fire. Montevidoni testified that within weeks. Rinderknecht became a target of arson investigators and that they began serving the first of more than 40 search warrants on Jan. 24.
Montevidoni said Rinderknecht refused to give investigators passwords for his devices and used a Tor browser. the gateway to the dark web. to find information on how to delete information from his iCloud as investigators closed in. Montevidoni testified that Rinderknecht told investigators he had a remote program running that would delete everything from his phone if federal authorities tried to use any password-cracking software.
Montevidoni said authorities were only able to conduct a partial extraction of his cellphone data.
In court, Rinderknecht has denied all wrongdoing.
As the trial continues, the government’s case rests on alleged actions and the digital trail it says supports them—while the defense keeps pointing back to what it says is missing: searches, prompts, and purchases that would show arson planning.
For the families and communities still dealing with the loss tied to the Palisades wildfire. the trial is playing out over fragments of a life captured in texts. searches. and screen recordings. At the center of it all is a question asked just minutes after the blaze erupted—one that may determine how jurors read everything that followed.
Palisades fire Lachman fire Jonathan Rinderknecht ATF ChatGPT arson trial Michael Montevidoni Steve Haney Luigi Mangione Tony Xu
So he asked ChatGPT about cigarettes and fires? Sounds like it was already in his head.
Wait I thought AI can’t be used like that? Like if he typed a question it doesn’t mean he did anything, right. Unless they’re saying ChatGPT told him to burn stuff which… come on.
They keep saying “burned underground for days” like that’s normal fire science. I don’t get how a cigarette even ties into re-igniting on Jan 7. Feels like prosecutors are reaching and using the AI query as the main smoking gun, but it’s not proof of arson by itself.
This is why I don’t trust AI. One question and now they’re like “yep arson.” Also Uber driver… I just feel like that matters for some reason. People act like computers are magic, but he still had to actually light the fire, so I’m not sure why they’re obsessing over the ChatGPT part.