Prosecutors' use of ChatGPT logs fails to sway jury in Palisades fire arson trial

Jury rejects ChatGPT logs as proof of arson intent

Prosecutors attempting to link Jonathan Rinderknecht to the January 1, 2025 Palisades wildfire—one of the deadliest in Los Angeles history—assembled a case combining location tracking, surveillance video, and witness accounts. According to The Verge AI, they also introduced a novel evidentiary element: transcripts and screen recordings from Rinderknecht’s interactions with OpenAI’s ChatGPT. However, the jury found the AI conversation logs unpersuasive, ultimately deadlocking at 10-2 in favor of the defense and prompting the judge to declare a mistrial.

Prosecutors’ ChatGPT evidence strategy

The prosecution’s digital-evidence approach centered on specific ChatGPT queries and requests. According to The Verge AI, Rinderknecht had asked the chatbot to generate images depicting fire, posed the question “Why am I so angry all the time?”, and vented frustrations about how wealthy individuals were damaging the world. Prosecutors also presented a screen recording in which Rinderknecht posed a hypothetical to ChatGPT: whether legal blame could attach to someone whose lit cigarette started a fire. The strategy appears to have rested on inferring malicious intent from patterns of inquiry and emotional expression recorded in the logs.

Juror skepticism signals emerging evidentiary limits

The mistrial outcome hinged partly on how a single juror—and likely others—viewed AI dialogue as evidence. According to The Verge AI, one juror told CBS LA that she routinely engages with ChatGPT herself and therefore did not regard Rinderknecht’s conversations as probative of criminal intent. She reported that the prosecution’s reliance on the logs actually provoked frustration, viewing it as an inappropriate leap from casual AI interaction to character inference. Her comment—“I talk to ChatGPT all the time”—underscores a fundamental divide: chatbot users perceive such systems as neutral thinking tools, not confessional spaces whose content reveals true character or intent.

Why This Matters

This mistrial sets an early precedent for how U.S. courts will evaluate AI-generated logs in criminal proceedings. Prosecutors face a methodological challenge: absent direct evidence linking a defendant to a crime, conversational data from AI systems may fail to establish motive or knowledge because jurors understand that people experiment, vent, and hypothesize with chatbots in ways they would not with human confidants. Defense counsel and civil-rights advocates will likely cite Rinderknecht’s hung jury as evidence that AI dialogue logs carry minimal probative weight without corroborating circumstantial or forensic evidence. If this trend holds across future cases, prosecutors may need to abandon AI logs as standalone indicators of intent and instead use them only to corroborate established facts (e.g., a ChatGPT timestamp placing a defendant in a specific mental state near the time of an alleged crime). The ruling also hints that courts may require explicit jury instructions distinguishing between typical AI use—which encompasses speculation, hypotheticals, and emotional expression—and actionable evidence of criminal planning.