A prompt on ChatGPT to produce the image of a burning city was a part of the evidence collected from the accused in the 2025 Palisade fire, which burnt more than 23,000 acres, destroyed nearly 7,000 structures, and killed 12 people in Los Angeles, California this January.Investigators have made three key references to information collected from the digital devices of Jonathan Rinderknecht, 29, specifically, his interactions with OpenAI’s pioneering large language model, which point to his involvement in starting the fire.Rinderknecht’s arrest is the latest based on evidence from interactions with artificial intelligence (AI), a novel digital avenue for investigators to glean through while working on a case. Just earlier this month, law enforcement in Florida sprang into action after a teen asked ChatGPT “how to kill my friend in the middle of class” on a school device.Rinderknecht’s fire obsessionAs per the indictment filed by investigators against Rinderknecht, he asked ChatGPT: “Are you at fault if a fire is lit [sic] because of your cigarettes?”. This, they believed, was because he wanted to “preserve evidence of himself trying to assist in the suppression of the fire”.“He wanted to create evidence regarding a more innocent explanation for the cause of the fire,” the indictment said.Prosecutors allege that Rinderknecht used ChatGPT to confide and indulge an obsession with fires.Two months before the fire, in November 2024, Rinderknecht had entered the following prompt on ChatGPT: “I am 28 years old. And… I basically… This just happened. Maybe like… I don’t know, maybe like 3 months ago or something. Like, the realization of all this. I literally burnt the Bible that I had. It felt amazing. I felt so liberated.”Story continues below this adThen on July 11, 2024, Rinderknecht told ChatGPT to produce a painting of a “burning forest” with a “crowd of people running away from the fire.His prompt: “A dystopian painting divided into distinct parts… On the far left, there is a burning forest. Next to it, a crowd of people is running away from the fire… In the middle, hundreds of thousands of people in poverty are trying to get past a gigantic gate with a big dollar sign… On the other side…a conglomerate of the richest people. They are chilling, watching the world burn down, and watching the people struggle…”How Palisades fire was startedRinderknecht was arrested in Florida on Tuesday (October 7). He has been charged with destruction of property by means of fire; officials might yet press further charges, including manslaughter.“The arrest, we hope, will offer a measure of justice to all those impacted,” Acting US Attorney Bill Essayli told a press conference on Wednesday in Los Angeles.Story continues below this adProsecutors alleged that Rinderknecht, an Uber driver who used to live in the Palisades, dropped off a passenger on New Year’s Eve before driving to a trailhead on the Temescal Ridge in the Pacific Palisades, just to the north of Santa Monica.Also Read | Southern California wildfires raging for over a week: Why firefighters are struggling to contain themThere he filmed videos on his iPhone, and “listened to a rap song — to which he had listened repeatedly in previous days – whose music video included things being lit on fire.” The blaze was allegedly triggered by a cigarette Rindereknecht had lit, although details remain unknown as yet.Prosecutors allege that Rinderknecht tried to call 911 after lighting the fire, but poor cell service hampered his call. A nearby resident informed authorities: the original blaze lit by Rinderknecht on New Year’s eve, known as the Lachman fire, was suppressed. An aerial view shows the devastation left by the Palisades Fire in the Pacific Palisades section of Los Angeles, January 27, 2025. (AP Photo)But it continued to smoulder deep underground, and flared up on January 7 during a thunderstorm. The fire would rage for nearly three weeks.Story continues below this adOpenAI and law enforcementIn a blog post in August, OpenAI said that when it detects users who are planning to harm others, it routes those conversations to specialised pipelines where they are reviewed by a small team trained on its usage policies and who are authorised to take action, including banning accounts.“If human reviewers determine that a case involves an imminent threat of serious physical harm to others, we may refer it to law enforcement,” OpenAI said.In Rinderknecht’s case, OpenAI, in a statement to Axios, said: “Following the Palisades fire tragedy, we responded to standard law enforcement requests related to this individual.” The statement added, however, that as yet, there is no evidence to suggest that the accused used ChatGPT to create content that was “against its policies”.Nonetheless, the case represents a watershed moment of AI intersecting with criminal law: Rinderknecht’s prosecution could shape both legal precedent and AI safety protocols in coming years.