A federal court prices an unchecked AI legal workflow: $3,000 sanctions each for the attorneys behind an AI-drafted brief with nearly thirty defective citations

The problem

Generative AI drafts legal prose that reads like law and cites like law — whether or not the law exists. The failure mode is invisible to a reader who does not check, and in February 2025 it reached a federal courtroom: defense counsel for Michael J. Lindell, FrankSpeech LLC, and My Pillow, Inc. in Eric Coomer’s defamation suit filed an opposition brief that had been run through AI and never citation-checked.

This is a verified negative: an independent federal court, in signed public orders, measured what that workflow produced and priced it.

What the court found

At the April 21, 2025 pretrial conference, Judge Nina Y. Wang identified “nearly thirty defective citations” in the 17-page brief. The defects, in the court’s words: “misquotes of cited cases; misrepresentations of principles of law associated with cited cases, including discussions of legal principles that simply do not appear within such decisions; misstatements regarding whether case law originated from a binding authority such as the United States Court of Appeals for the Tenth Circuit; misattributions of case law to this District; and most egregiously, citation of cases that do not exist.”

The colloquy that followed is the record’s core artifact:

THE COURT: Was this motion generated by generative artificial intelligence? MR. KACHOUROFF: Not initially. Initially, I did an outline for myself, and I drafted a motion, and then we ran it through AI. THE COURT: And did you double-check any of these citations once it was run through artificial intelligence? MR. KACHOUROFF: Your Honor, I personally did not check it. I am responsible for it not being checked.

Counsel’s later explanation — that an earlier draft had been filed by mistake — did not persuade the court, which noted the contradictory statements and lack of corroborating evidence.

The order

On July 7, 2025, the court made its Order to Show Cause absolute and found that both attorneys “violated Rule 11 because they were not reasonable in certifying that the claims, defenses, and other legal contentions” in the brief “were warranted by existing law or by a nonfrivolous argument for extending, modifying, or reversing existing law or for establishing new law.”

The sanction: $3,000 against Kachouroff and his firm jointly and severally, and $3,000 against DeMaster individually — “the least severe sanction adequate to deter and punish defense counsel in this instance” — payable to the court’s Registry by August 4, 2025. The court wrote that it “derives no joy from sanctioning attorneys who appear before it.”

Scope matters: the sanctions landed on counsel only. Because Kachouroff confirmed he had not told his clients he uses AI tools in his practice, the court did not extend the sanctions to Lindell, FrankSpeech, or My Pillow.

Why this matters for AI-led operations

Every figure in this story originates with an independent measurer — a federal judge with no stake in AI vendors or in the litigants’ marketing. That makes it one of the rare AI-outcome records whose critical numbers do not come from the party that deployed the system. The lesson it documents is operational, not legal: an AI step inserted into a professional workflow without a verification step behind it produced ~30 defects in one 17-page document, and the cost was borne by the professionals who signed the output.

Verification record

• Status: pending — desk research complete and checker-verified; all critical figures sit in court-signed public orders. Awaits human sign-off (not verified).
• Primary sources: Order to Show Cause [Doc. 309, 2025-04-23] and sanctions order [Doc. 383, 2025-07-07], D. Colo. No. 1:22-cv-01129-NYW-SBP, fetched from the public RECAP archive and byte-matched to same-day Wayback captures.
• Independent relays: NPR (2025-07-10), The Colorado Sun (2025-07-07).
• Not claimed: the exact underlying jury-verdict figure (artifact not yet fetched; NPR’s “more than $2 million” is used); the May 2026 follow-on $5,000 sanction (separate later order, not yet fetched).