The cover-up is worse than the citation: what courts punish hardest

Ask any litigator and you will hear the same thing: the court can forgive the mistake. It does not forgive being misled about it.

The AI cases show this clearly.

Couvrette v. Wisnovsky (D. Or.)

This is the case to point to when someone asks how bad it can get. The total came to roughly $110,000, and the case was dismissed with prejudice. The briefs had 15 references to nonexistent cases and eight fabricated quotations. When the other side flagged them, the lawyer removed the material and refiled. The judge called that an attempted cover-up and found the amended briefs still misstated the law. The fees and fines followed.

Fletcher v. Experian (5th Cir., Feb. 2026)

The Fifth Circuit imposed $2,500. That is small. What matters is the reasoning. The court found the lawyer's responses to its show-cause order evasive and misleading. It also declined to adopt a local rule requiring AI certification, on the view that Rule 11 already requires accurate filings, and made the point that ignorance of AI's risks is no longer a defense. You do not need a new rule to be held to account.

Coomer v. Lindell (D. Colo.)

The court sanctioned counsel $3,000 each in 2025. In May 2026 it imposed another $5,000. A second round means the first did not end the problem. The court kept the file open and kept looking.

A caution on one more case

Whiting v. City of Athens (6th Cir., March 2026) is often listed in these roundups with a total of $30,000. The court did not find that AI was used. If you cite it, say that.

What the response looks like when it goes right

Across these orders, courts reward the same three things:

1. Prompt, plain disclosure. Tell the court what happened when you find out.
2. A specific account. What was generated, by what, who reviewed it, what was missed.
3. A real fix. A change in process, not a promise to be careful.

The firm in Johnson v. Dunn did that. It audited over 2,400 citations after the problem surfaced. The firm was not sanctioned. Individuals were.

The practical problem

Most lawyers cannot give a court a specific account, because they do not know exactly what happened. The AI was used in a separate tool, in a separate window, with no record. When the judge asks "how did this get into your brief," the honest answer is "I'm not sure," and that is the answer that tempts people to say something less honest.

A record changes that conversation.

In JILL, each prompt, retrieval and output is recorded in an append-only, cryptographically signed audit log under DDEAS, chained so it is tamper-evident. If a citation is challenged, you can show what JILL produced, what it was validated against and what confidence it carried. If the problem turns out to be your own edit after the fact, the record shows that too.

That is not about hiding anything. It is the opposite. It lets you answer a court's question with facts instead of memory.

What to do before it happens

• Decide now who is responsible for verifying citations on every filing.
• Use tools that validate authority before it reaches a draft.
• Keep the record of how AI-assisted work was produced.
• If something slips through, say so immediately and say exactly what happened.

The sanction for the mistake is a cost. The sanction for the cover-up is a verdict on your credibility. Courts have shown which one they weigh more heavily.

Learn about the audit trail →