Gauthier v. Goodyear is the case where a lawyer who said he checked AI output with a second AI tool was still sanctioned. In November 2024, a federal judge in the Eastern District of Texas sanctioned attorney Brandon Monk $2,000 for filing a brief with fabricated citations after he used one AI tool to generate case law and a second AI tool to verify it. Neither caught the hallucinations.


Background

The underlying case was a wrongful termination dispute. An employee sued Goodyear Tire & Rubber Co. in the Eastern District of Texas, and the case reached the summary judgment stage. Attorney Brandon Monk was tasked with drafting a brief opposing Goodyear's motion for summary judgment.

Another lawyer-and-AI citation matter is covered in People v. Crabill.

Monk used an AI tool to generate the legal research and arguments for his opposition brief. The AI produced case citations and legal arguments that appeared to support his client's position. But rather than verifying the output in a traditional legal database, Monk took what he considered a modern approach: he ran the AI-generated content through a second AI product to check for accuracy.

The second AI tool didn't flag any problems. The citations passed what amounted to an AI-on-AI quality check. Monk filed the brief with the court, confident that the dual-AI workflow had caught any errors. It hadn't.

Gauthier v. Goodyear Tire & Rubber Co.
No. 1:23-cv-281 (E.D. Tex. 2024)
Court
U.S. District Court, Eastern District of Texas
Date
2024-11-25
Category
AI Hallucination / Sanctions
Sanctions
$2,000
AI Case Law: Updated April 2026

What Happened

The brief contained fabricated citations and legal arguments. The cases Monk cited didn't exist. Judge Marcia A. Crone identified the problem and ordered Monk to explain.

Monk admitted to the court that he had used one AI tool to produce the response and relied on another AI product to verify it. The order records that he submitted the response without reading the cases or confirming they existed, and that he first tried to locate them after the show cause order, although Goodyear's reply brief had already flagged the problem. The only check he described was a Lexis AI feature, which did not flag the issues.

The court was unimpressed. Judge Crone found that Monk failed to make the reasonable inquiry Rule 11 requires into the validity of his arguments, and that he did not correct them when Goodyear identified the problem. The Lexis AI check he described did not change the outcome: the citations were fake, and no human had verified them.


The Ruling

Judge Crone ordered Monk to pay a $2,000 penalty to the court registry. She also required him to complete a continuing legal education course on AI in legal practice.

The order does not rule on AI-to-AI checking as such. It holds that Monk did not make the reasonable inquiry Rule 11 requires: he filed the response without reading the cases or confirming they existed and did not correct it when Goodyear flagged the problem. That a second AI tool is no substitute for checking the cases yourself is our reading, not the court's holding.

The CLE requirement was notable. It signaled that the court viewed Monk's conduct as stemming partly from a lack of understanding about AI limitations. The penalty was lighter than some other AI sanctions cases, likely because Monk was cooperative and transparent about his process. But the legal principle it established was significant.

Outcome: Attorney Monk was ordered to pay a $2,000 penalty to the court registry and required to complete a continuing legal education course on AI in legal practice.

Why This Case Matters

In Gauthier v. Goodyear the attorney said he drafted with one AI tool and tried to check the result with a Lexis AI feature, which did not flag the fake citations. The order sanctioned him for filing without reading the cases or confirming they existed.

The ruling matters because it reflects how AI tools actually work. Large language models don't independently verify facts. They generate plausible text. Running hallucinated content through a second LLM doesn't fact-check it. The second model simply evaluates whether the text looks plausible, which it does, because the first model generated it to look plausible. It's a circular validation loop.

This case also pushed the conversation toward what adequate verification actually looks like. The court did not address AI checking AI as such: its finding was that Monk never read the cases or confirmed they existed. Our reading is that human review against authoritative sources is the floor, not the ceiling, for AI-assisted legal research.


Lessons for Attorneys

There's no AI shortcut for citation verification. Whether you use one AI tool or five, the verification step must involve a human checking against an authoritative legal database. Westlaw, Lexis, Fastcase, Google Scholar, PACER. Pull up the actual case. Read the actual opinion. Confirm it says what your brief claims it says. This is non-negotiable.

Understand why AI-on-AI checking fails. Large language models assess plausibility, not accuracy. A second model will confirm that a fabricated citation looks like a real citation because the first model was specifically designed to generate realistic-looking text. The second model has no independent access to a case database. It's evaluating form, not substance.

If your firm is building AI workflows, build the human verification step into the process as a hard gate, not an optional review. The brief doesn't move to the next stage until every citation has been manually confirmed. Document the verification. Keep a log of which citations were checked, by whom, and against what source. This creates a defensible record if a citation issue ever arises.


The Bottom Line

AI can't verify AI. In Gauthier v. Goodyear the attorney's Lexis AI check did not catch the fake citations, and the court sanctioned him for not reading the cases or confirming they existed. Human review against authoritative legal databases is the only acceptable standard.

AI-Assisted Research. This piece was researched and written with AI assistance, reviewed and edited by Manu Ayala. For deeper takes and the perspective behind the research, follow me on LinkedIn or email me directly.

Price sources, checked 22 October 2026: Gavel. Each figure above is a list price from the vendor's own pricing page, or arithmetic on one. Plans, seat minimums and promotions change; confirm on the vendor's page before you buy.