Prosecutors in the Palisades wildfire arson trial introduced Jonathan Rinderknecht's ChatGPT conversation logs as evidence against him, and on June 28, 2026 a Los Angeles jury voted 10-2 in favor of the defense, forcing the judge to declare a mistrial. Rinderknecht was charged with setting the New Year's Day 2025 fire that became one of the deadliest wildfires in Los Angeles history. The chatbot transcripts sat alongside iPhone location data, security camera footage, and witness testimony in the state's case.
The ChatGPT evidence was unusually personal. Prosecutors told the court Rinderknecht had used the chatbot to generate images of fire, vented to it about the wealthy destroying the world, and at one point typed a question about his own emotional state.
That question, according to the prosecution's submission, was direct.
“Why am I so angry all the time?”— Jonathan Rinderknecht, Defendant in the Palisades fire trial
Key facts
- 01Prosecutors introduced Jonathan Rinderknecht's ChatGPT logs as evidence in the Palisades wildfire arson trial.
- 02The jury voted 10-2 in favor of the defense, leading the judge to declare a hung jury and mistrial on June 28, 2026.
- 03Rinderknecht allegedly asked ChatGPT to generate images of fire and queried whether a cigarette-lit fire could be blamed on the smoker.
- 04Prosecutors also relied on iPhone location data, security camera footage, and witness testimony alongside the chatbot logs.
- 05One juror told CBS LA the ChatGPT evidence made her 'angry' because she uses the chatbot herself.
A separate screen recording shown to jurors captured Rinderknecht asking ChatGPT whether someone could be blamed for a fire if it was lit by their cigarette. Prosecutors framed the query as consciousness of guilt, an attempt to test a legal theory with an AI assistant in the days surrounding the blaze. The defense countered that idle chatbot queries are not evidence of intent, and the jury appears to have agreed.
The 10-2 split tilted toward acquittal rather than conviction, which is what produced the hung jury. In California a unanimous verdict is required in a criminal trial, and the lopsided lean toward the defense suggests the prosecution's narrative — including its reliance on the ChatGPT logs — failed to land with most of the room.
One juror, speaking to CBS LA after the verdict, said she found the chatbot evidence unpersuasive and personally offensive.
“I talk to ChatGPT all the time.”— Anonymous juror, Juror speaking to CBS LA
She added that she was angered by the suggestion that talking to a chatbot indicated any kind of character flaw, a reaction that points to how normalized consumer AI use has become with juries drawn from the general public. What prosecutors treated as suspicious, at least one juror treated as Tuesday.
The case is one of the first high-profile US criminal trials in which a defendant's conversations with a general-purpose AI assistant were entered as substantive evidence. OpenAI was not a party to the proceedings, and the logs were obtained through legal process directed at the user's account rather than at the company's broader user base. Still, the trial sets a practical precedent: prosecutors are now treating chatbot transcripts the way they have long treated text messages, browser history, and search queries.
The defense's success in the Rinderknecht trial does not settle the legal question. A retrial is possible, and other prosecutors will continue to subpoena chatbot logs in serious cases — particularly when the queries appear to map onto the alleged crime. Defense attorneys will increasingly need to anticipate AI-conversation evidence and prepare jurors to read it in context rather than as confession.
The juror reaction matters more than the verdict itself. Prosecutors built a theory around the idea that asking a chatbot about anger, fire, and culpability was probative of arson. A jury raised on Google search history accepted that framing for queries typed into a search bar; a jury that uses ChatGPT for therapy, meal plans, and venting did not extend the same logic to a conversational AI. For OpenAI and its competitors, that gap is now part of the product's social meaning. For prosecutors, it is a new evidentiary problem — one where the more useful a chatbot becomes in daily life, the weaker any single transcript looks as proof of a crime.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
