Company case

Butler Snow: a firm with an AI policy, and a partner who did not follow it

The firm’s policy required written approval before using ChatGPT for legal research. A partner used it without approval and filed five citations it had made up. The court removed three lawyers from the case. The firm paid for a check of 2,400 citations.

Outcome Went wrong

Something failed in work where AI was used, and there is a public record of it.

In the guide What fails

At a glance

Company
Butler SnowLaw firm
Size
Mid-market (250–999 employees)
Industry
Professional services
Where and when
United States (federal court in Alabama). May – July 2025.
Who produced the figures
Court or tribunal recordThe facts, figures and quotations come from the court’s published sanctions order, which quotes the firm’s own statements to the court.

What happened

The firm was defending a former state prison official. In May 2025 its lawyers filed two motions. A partner had revised parts of them. To find supporting case law he ran a search in ChatGPT and put the results in without checking them.

The other side pointed out that the citations looked made up. The lawyers admitted it in writing the same day the court asked.

The numbers

  • Five citations across two motions were wrong. ChatGPT had made them up.
  • Three lawyers were publicly reprimanded, removed from the case and referred to their licensing bodies. Two more were released without penalty, and so was the firm.
  • 52 court files were reviewed by the firm itself for the same problem.
  • More than 2,400 citations in 330 filings were then checked by 28 lawyers from another firm, at Butler Snow’s expense. They found no other made-up citation.

The policy that was already there

The firm had rules. Its practice group leader told the court:

Under firm policy, the use of ChatGPT for legal research requires written approval from a practice group leader.

He said the firm had made the limits on AI “abundantly clear to all of our attorneys”. The firm’s guidance also said that all AI output must be reviewed and verified by the responsible lawyer before it is filed or relied on.

The partner who used ChatGPT was the assistant leader of that same practice group. He told the court he knew about the limits and “did not comply with the policy”.

What the firm changed

  • It reminded all its lawyers of their duty to verify every citation.
  • It promised firm-wide training on the proper use of AI.
  • It said it would adopt a check before every court filing: review all the legal sources in the document, and confirm that each one exists, is accurate and is relevant.

What the court said

The judge wrote that fines and public embarrassment had not stopped lawyers filing made-up cases, so she chose a stronger penalty: removal from the case. She did not suspend the lawyers from practice. She said the firm’s own review and the outside check reassured her that this was not needed.

What the case shows

  • A written policy did not stop a senior person. The rule existed and was known. What was missing was a check that did not depend on anyone remembering the rule.
  • The firm’s answer was a step in the process. A check before every filing works even when someone ignores the policy.
  • Cleaning up cost far more than checking would have. Five citations led to a review of more than 2,400.
  • Fast, open correction counted. The court said so when it explained why the penalty went no further.

What this says about the practices

Practices the company applied

Practices that were missing

What the company changed afterwards

Sources

  1. Johnson v. Dunn, No. 2:21-cv-1701-AMM. Sanctions order, 23 July 2025. US District Court, Northern District of Alabama. storage.courtlistener.com