Prosecutors used a man's ChatGPT logs to try to link him to the Pacific Palisades fire, but it didn't work

midian182

Posts: 11,718   +177
Staff member
In a nutshell: Users have long been warned to be careful what they tell ChatGPT, especially if it could get them in trouble with the law. California prosecutors, for example, used the chatbot's logs in their case against the man accused of starting a blaze that became the Pacific Palisades fire of 2025.

Jonathan Rinderknecht was accused of deliberately starting last year's fire, which killed a total of 31 people and destroyed more than 6,000 buildings.

The prosecution's case included everything from his iPhone's location data and security camera footage to witness testimonies.

Prosecutors also used Rinderknecht's ChatGPT logs. They said he used the tool to generate an image depicting a city on fire, asked the AI "why am I so angry all the time?" and went on tirades about how the rich were destroying the world – allegedly influencing his decision to start the blaze in a wealthy neighborhood.

Rinderknecht is said to have used ChatGPT like a diary. One prompt asked if he could be held responsible for a fire that was caused by his fallen cigarette.

Prosecutors said that Rinderknecht was motivated by revenge, anger, and loneliness because he had nowhere to go for the holiday.

The trial against Rinderknecht ended last week. After two days of deliberation, the jury voted 10-2 in favor of the defense, leading the judge to declare a hung jury and a mistrial.

Defence attorney Steve Haney argued that Rinderknecht – who has both French and US citizenship and is a former Uber driver – had not started the blaze and had called authorities after seeing the fire break out.

While the ChatGPT evidence might appear damning, a juror told CBS LA that she didn't believe the logs proved anything, noting that she talked to the bot all the time. She added that it made her "angry" that Rinderknecht's use of ChatGPT was being framed as a character flaw.

The decision doesn't mean the case against Rinderknecht is over. A retrial has been scheduled for October this year, with Rinderknecht remaining in custody until then.

This isn't the first time that ChatGPT has been caught up in a criminal investigation. Florida Attorney General James Uthmeier announced in April that the Office of Statewide Prosecution launched a criminal investigation into OpenAI and ChatGPT over its role in a school shooting last year.

"If ChatGPT were a person, it would be facing charges for murder," Uthmeier said in a previous news release.

Permalink to story:

 
I imagine future wars will see much of the insanely expensive datacenters being primary targets, along with energy infrastructure as we see in the current Ukraine v Russia war.
 
I imagine future wars will see much of the insanely expensive datacenters being primary targets, along with energy infrastructure as we see in the current Ukraine v Russia war.

While massive data centers will likely remain vulnerable primary targets for training models, I actually think the operational reality of warfare will force AI deployment to decentralize.

Just as Arpanet was designed to route around outages, military AI must route around the destruction of infrastructure.

I think we’re more likely to see a shift that appears similar to the PC revolution with advances driven more toward miniaturization and energy efficiency. A future belonging more to 'Edge AI'—localized, low-power hardware embedded directly into mobile, autonomous systems that can mesh together to scale compute or operate entirely in isolation.

By eliminating single points of failure and cloud dependencies, distributed AI networks would be highly resilient against localized grid failures and communication jamming, making them significantly harder to disrupt.
 
If he didn’t start the fire, then can we start blaming the state of California for not doing anything to prevent it?
Yes, California and LA specifically are the real culprits here. Major forest fires are 100% preventable disasters: clear dead and dying trees, cut firebreaks, remove brush buildup ... all things the enviro-Nazi Left refuses to allow. After the Pacific Palisades fire, we disovered some of the areas impacted hadn't had brush cleared in several decades.
 
While massive data centers will likely remain vulnerable primary targets for training models, I actually think the operational reality of warfare will force AI deployment to decentralize.

Just as Arpanet was designed to route around outages, military AI must route around the destruction of infrastructure.

I think we’re more likely to see a shift that appears similar to the PC revolution with advances driven more toward miniaturization and energy efficiency. A future belonging more to 'Edge AI'—localized, low-power hardware embedded directly into mobile, autonomous systems that can mesh together to scale compute or operate entirely in isolation.

By eliminating single points of failure and cloud dependencies, distributed AI networks would be highly resilient against localized grid failures and communication jamming, making them significantly harder to disrupt.
Only 500 powerful drones could take the decentralized system down. The sites are massive, making them easy targets for disruptions. Otherwise, you're correct.
 
If he is convicted in the new trial, "most" of the blame should be on him, but 45% of it should be on the people of California & the politicians they elected, that would not allow clearing out of the underbrush and what not. Not to mention the LACK of water thanks to mayor Karen Bass.
 
Back