A federal trial that drew unusual attention for its evidence — a defendant's chats with an AI assistant — has ended without a verdict. Jurors deadlocked in the arson case against Jonathan Rinderknecht, and a judge declared a mistrial, the technology site The Verge reported.
The case
Rinderknecht is accused of igniting a blaze that, days later, grew into the Palisades Fire — one of the most destructive wildfires in Los Angeles history, part of the January 2025 firestorms that killed people and destroyed thousands of structures. He faces federal arson charges and has denied wrongdoing; he is presumed innocent. On Friday, June 26, after the jury split 10 to 2 in favor of acquittal, U.S. District Judge Anne Hwang declared a mistrial, according to ABC7 Los Angeles. The judge set a tentative retrial for October and ordered Rinderknecht to remain in custody.
How ChatGPT entered the courtroom
Prosecutors told the jury that Rinderknecht's exchanges with ChatGPT revealed a state of mind, arguing he was angry, fixated on wealth disparity and seeking "revenge" against society. The most-discussed piece of evidence was a screen recording from his phone that, prosecutors said, showed him dialing 911 about the fire and then asking ChatGPT whether a person could be at fault if a fire was started by their cigarette. Notably, prosecutors did not frame that as a confession: they argued it was an attempt to manufacture an innocent explanation, as CNN reported. The defense countered that he had gone to a hilltop to watch fireworks, that the initial fire was sparked by fireworks rather than arson, and that calling 911 showed he tried to help — and the jury's lean toward acquittal suggested many found the AI evidence less than conclusive.
Why it matters beyond this trial
The case is part of a broader shift: law-enforcement agencies increasingly treat conversations with AI chatbots as a potential source of evidence, obtained from companies like OpenAI through warrants or subpoenas. Unlike a conversation with a lawyer, a chat with an AI assistant carries no legal privilege, and users may not realize their queries can be retrieved and used in court. Civil-liberties advocates have warned that as people pour ever more of their thoughts into AI tools, those logs become a rich and largely unprotected record — one that, as the Palisades case shows, prosecutors are now willing to put in front of a jury.
What happens next
With the mistrial declared, the matter is not over: prosecutors can retry the case, and a new trial was tentatively scheduled for October, with a status hearing set for July. Whatever the eventual outcome for Rinderknecht, the trial has already become an early marker in a question courts will face more and more often — how much weight a person's words to a machine should carry when their liberty is at stake.
