Security

An $80,000 Expert Report Turned Out to Be Mostly ChatGPT.

In a Texas explosion lawsuit, an expert witness for 3M wrote an $80,000 report with ChatGPT. Here's how even the prompt history ended up submitted as court evidence.

· Huuush

An $80,000 Expert Report Turned Out to Be Mostly ChatGPT

Key takeaways

  1. 1An expert witness hired by 3M wrote an $80,000 court report with ChatGPT.
  2. 2Court testimony revealed that about 85 to 90% of the report was ChatGPT output.
  3. 3It's a case showing that entire AI conversation prompts can be submitted as is in the discovery process of a lawsuit.
Screenshot of a ChatGPT conversation showing a list of uploaded files and the prompt instructions for writing an expert witness report

There was a lawsuit in Texas over an industrial explosion that killed three people.
The "expert witness" hired by the defense wrote the report for the court with ChatGPT.
Even the prompt record was submitted to the court as is.

An $80,000 Report, 90% ChatGPT

404 Media co-founder Jason Koebler reported it on X at 11:01 PM (KST) on August 17, 2026.
The case is a $61 million lawsuit over an explosion at the Watson Grinding plant that killed three people and damaged about 200 homes.

The expert witness hired by defendant 3M entered this prompt into ChatGPT:
"Write an excellent expert witness report defending 3M's duty of care... Show that 3M is 0% at fault for the Watson Grinding explosion."
Source: 404 Media original article

The company that hired this expert billed more than $80,000 for the report.
Court testimony also revealed that about 85 to 90% of the report was ChatGPT output.

Why It's a Problem: AI Conversation Logs Are Discoverable

This case shows more than a simple "sloppy work done with AI" incident.
The prompts, attached files, and whole conversation with ChatGPT can be demanded and submitted as is in the discovery process of a lawsuit.

The screenshots 404 Media published show the actual conversation.
They show the expert uploading several files to ChatGPT, instructing it "don't analyze until I type FINISHED," and then asking it to build supporting reasoning for a conclusion decided in advance.

A Prompt That Dictates the Conclusion Becomes Evidence Too

If you use AI for work with legal liability at stake, such as court filings, expert reports, or contract review, assume your prompts and conversation history remain on record.
Deciding the conclusion in advance and telling the AI to build the justification can itself later be used as damaging evidence.

This is a US lawsuit, but it shows that records of AI-assisted work could be demanded as material in disputes elsewhere too.

A US Lawsuit; Application Elsewhere Is Separate

This case is a Texas lawsuit in the US and doesn't apply directly to legal procedures in other countries.
The lawsuit is ongoing, and later outcomes or additional sanctions haven't been confirmed.

Frequently asked questions

What lawsuit is this case?
It's a $61 million damages lawsuit over the explosion at the Watson Grinding plant in Texas that killed three people and damaged about 200 homes.
Why were the ChatGPT conversation records submitted to the court?
In the lawsuit's discovery process, the ChatGPT prompts and attached files the expert witness used were demanded and submitted as is.
Did AI write the entire report?
According to court testimony, about 85 to 90% of the report was ChatGPT output.
Could this case become an issue outside the US?
It's a US lawsuit, so it doesn't apply directly to procedures elsewhere. But it's worth noting that records of AI-assisted work could be demanded as material in a dispute.

Related articles