ChatGPT Conversations Used as Evidence for the First Time in a Deadly Arson Case

A user's ChatGPT conversations have been officially used as evidence in a US criminal court for the first time, in an arson case tied to one of the deadliest wildfires in Los Angeles history. The trial ended in a mistrial, but the precedent is already set.

The Palisades fire case

Jonathan Rinderknecht was charged with deliberate arson on 1 January 2025 in a hillside area of Los Angeles. A small initial blaze grew into a catastrophe within hours: thousands of homes destroyed, several people killed, tens of thousands of residents evacuated. Damage is estimated in the billions of dollars, making the Palisades fire one of the most destructive in California history.

Prosecutors built their case on conventional evidence: iPhone geolocation data placed Rinderknecht at the site of the ignition, surveillance cameras showed his movements, and witnesses confirmed his presence. But one unusual element was added to all of this: his ChatGPT conversation history.

What the logs showed

According to the prosecution, Rinderknecht used the chatbot heavily in the days before the fire. Prosecutors presented the conversations to the court as evidence of intent:

• Requests to generate images of fire and burning objects
• A personal question to the chatbot: "Why am I always so angry?"
• Extended musings on how wealthy people "are destroying the world"
• A screen recording in which the defendant asked ChatGPT whether a person can be held responsible for a fire if they started it themselves

Prosecutors were especially interested in the last question. In their account, it shows that Rinderknecht was weighing the legal consequences of his actions before carrying them out, meaning he acted deliberately.

A new kind of digital evidence

As far as is known, this is one of the first cases in the US in which the history of conversations with an AI chatbot was formally presented in court as evidence. Previously, digital evidence involved messaging apps, email, or search history. ChatGPT adds a fundamentally new type of data to that list.

The difference is significant: people tell a chatbot things they would not type into a search engine, such as personal troubles, fears, and plans. That candor makes the logs a sensitive source for investigators.

OpenAI cooperates with law enforcement when presented with a court order, and most users never think about this. An option to turn off chat history exists, but conversations are recorded by default.

"People talk to

ChatGPT as if it were a diary or a therapist, without considering that these records are stored and can be requested by investigators," warn civil rights advocates.

Despite all this, the trial ended in a mistrial: the jury could not reach a unanimous decision. A retrial is possible, but the legal weight of ChatGPT logs as evidence of intent has not been officially established.

What this means

The Rinderknecht case sends a clear signal: anything written to an AI chatbot can potentially end up in the materials of a criminal case. For users, it is a reason to review their privacy settings and the option to turn off chat history. For lawyers and AI companies, it signals that the legal status of conversations with AI needs serious thought: the precedent is set, and the next trial may cement it for good.