The Palisades area of Los Angeles was hit by a severe wildfire on New Year's Day 2025, becoming one of the deadliest wildfires in the city's history. The authorities charged Jonathan Rinderknecht with arson. During the trial, they obtained the location data of his iPhone, surrounding surveillance videos, and witness statements in an effort to prove that he was present near the fire and participated in the arson. Along with traditional evidence, records of his chats and image generation using ChatGPT were presented to the court, which is still rare in criminal cases.
According to prosecutors at trial, Lindknecht asked ChatGPT to generate images related to "fire" and complained to the chatbot about "why he was always so angry." Additionally, he ranted in the conversation, accusing wealthy people of "ruining the world," and was noted to have asked ChatGPT if a fire was started by someone's cigarette, if that person would be held accountable. The prosecutor tried to use this to shape the defendant's mental state and motivation, implying that he had an abnormal focus on flames and had strong resentment towards social injustice.
However, this argument, built around ChatGPT chat records, did not convince the jury. The jury reportedly voted 10-2 in favor of the defense in a final vote, leaving the verdict deadlocked. In the U.S. criminal justice system, major cases often require juries to reach a unanimous or near-unanimous verdict, and a deadlock means the inability to reach a clear verdict of guilt or innocence. The judge then declared that the jury was "divided," ruling it a "jury deadlock" in accordance with the law and declaring a mistrial.
One juror told local outlet CBS LA that she didn't think the ChatGPT records proved anything substantial. She confessed that she "chatted a lot with ChatGPT," and found the prosecution's attempt to infer character flaws from the chat logs "irritating." In her view, interacting with chatbots has become a part of ordinary people's daily lives, and these logs alone are difficult to use as a basis for judging whether a person has criminal tendencies.
This case reflects new controversies over the use of artificial intelligence tools in judicial scenarios: On the one hand, prosecutors and investigative agencies are increasingly inclined to consider personal device data, online activity records, and even interaction content with AI as potential sources of evidence; on the other hand, the public and jurors still have strong doubts about how to interpret these data and whether they adequately reflect true motives and behaviors. In the Palisades wildfire case, ChatGPT records failed to hold the key to a conviction and instead stoked resentment among jurors about the prosecution's evidentiary strategy.
As generative AI becomes increasingly integrated into everyday use, similar disputes are likely to be repeated in more cases in the future. How the judicial system distinguishes "daily chat" from "danger flags" and how to avoid generalized stigmatization of technology users will become practical issues that need urgent response at the intersection of law and technology.