FOLLOW US

What the Court Just Changed

• A federal judge dismissed a USDA employee’s 67-page discrimination complaint after saying the filing appeared to have been generated by artificial intelligence.
• The court did not establish that AI actually wrote the complaint, but ordered the plaintiff to personally draft her amended pleading without AI assistance.
• The ruling goes beyond fake AI citations and raises a harder question—when can a judge identify AI use simply from the way a legal document is written?

By Samuel López | USA Herald

ARIZONA - A federal judge in Arizona has given a government employee an unusual instruction as she tries to keep her discrimination lawsuit alive… rewrite the complaint yourself, and do not use artificial intelligence to draft it.

U.S. District Judge Krissa M. Lanham dismissed Shelly George’s lawsuit against Agriculture Secretary Brooke Rollins after finding that George’s 67-page, 199-paragraph complaint violated federal pleading requirements.

The October 6 order in George v. Rollins, No. CV-26-08218-PCT-KML, said the complaint “appears to have been generated by artificial intelligence” because it resembled other AI-generated complaints the court had encountered.

But that wording matters.

The court did not identify ChatGPT, Claude, Gemini or another AI system as the author. Nor did the order establish through metadata, an admission or other direct evidence that George had used a particular AI tool.

Instead, the judge pointed to characteristics of the filing itself.

The complaint contained extensive factual allegations that were difficult to follow and repetitive, discussed legal authorities, referenced exhibits that were not attached and shifted between theories involving retaliation, hostile work environment and what the complaint called a retaliatory hostile work environment.

The court classified the filing as a prohibited “shotgun pleading.”

That legal problem exists independently of artificial intelligence.

Federal Rule of Civil Procedure 8 requires a complaint to contain a “short and plain statement” showing why the plaintiff is entitled to relief. Earlier this year, the Ninth Circuit explained that courts may dismiss shotgun pleadings when allegations are vague, repetitive or fail to connect particular facts with particular legal claims.

That distinction is important because George’s case was not dismissed merely because a judge disliked AI.

It was dismissed because the complaint, in the court’s view, failed to perform one of the most basic functions of a federal pleading—telling the defendant what claims are being brought and which facts support them.

George was given another chance.

Judge Lanham ordered her to file an amended complaint of no more than 25 pages by October 19. This time, George must personally draft the allegations she believes are relevant and identify the claims she intends to pursue. The court expressly prohibited her from using AI tools to draft that amended complaint. If she does not file it, the court directed that the action be dismissed with prejudice.

There is another piece of the order that deserves more attention.

The judge also requires parties using generative AI in filings before her to disclose that use and certify that the material has been personally reviewed and verified.

That means the court is drawing a line between AI assistance and AI substitution.

A litigant may use technology in some circumstances, but the human filing the document remains responsible for what reaches the docket.

And George’s case presents a particularly interesting problem for courts - what happens when there is no fabricated case citation or obvious hallucination?

Many earlier AI controversies involved lawyers submitting nonexistent cases. Here, the warning signs identified by the court were structural—excessive length, repetition, disconnected facts and unclear legal claims.

In other words, the alleged AI fingerprint was the writing itself.

That creates a new problem as AI becomes routine in legal work. Poor human drafting can be repetitive. AI drafting can be excellent. And increasingly, distinguishing one from the other by style alone may become difficult.

The safer standard may ultimately have little to do with whether AI touched a document at all.

The question courts appear increasingly interested in is simpler; Who checked it, who stands behind it, and can the person filing it explain every allegation and legal argument on the page?

About the Author

Samuel López is a reporter and news correspondent for USA Herald with more than two decades of experience working in the legal sector. His background includes analyzing state and federal litigation, complex pleadings, discovery, motions and appellate matters. His reporting frequently examines the growing collision between artificial intelligence, legal ethics, court procedure and the responsibility lawyers and litigants retain when technology is used to produce legal work.

Keep Reading Beyond the Headlines

If you found this report useful, join readers who follow USA Herald for independent legal reporting, emerging AI issues, national-security developments and stories that often receive little attention elsewhere. Subscribe through the USA Herald newsletter and stay connected as these cases develop.

👉 Follow USA Herald on X at https://x.com/RealUSAHerald