Johnson v. Dunn: three lawyers are disqualified and referred to the bar for filing five ChatGPT-hallucinated citations

The problem

In Johnson v. Dunn (No. 2:21-cv-1701-AMM, N.D. Ala.), a prisoner civil-rights case, incarcerated plaintiff Frankie Johnson accused defendant Jefferson Dunn, the former Commissioner of the Alabama Department of Corrections, of citing fabricated legal authorities in two motions (source). Those two motions had been filed by three of Dunn’s attorneys at Butler Snow LLP (source). The court ordered the attorneys to show cause, and their own account made the cause plain (source).

What was built

Three attorneys for Dunn, Matthew B. Reeves, William J. Cranford, and William R. Lunsford, confirmed that the citations were “hallucinations of a popular generative artificial intelligence (‘AI’) application, ChatGPT” (source). In the court’s words, “the citations were completely made up” (source). The opinion records that “in total, there were five problematic citations across two motions” (source). An independent legal-industry account states that Reeves “admitted he used ChatGPT to generate legal citations for both motions and inserted them without verifying a single one” (source). A separate newsroom, reporting the Associated Press wire, quotes Judge Manasco describing the conduct as “recklessness in the extreme” (source).

The outcome

On July 23, 2025, Judge Anna M. Manasco held that fabricating legal authority “demands substantially greater accountability than the reprimands and modest fines that have become common as courts confront this form of AI misuse” (source).

The count that gets quoted: five fabricated citations across two motions. The court itemized each of the five, beginning with a fake United States v. Baker cite offered as “confirming broad discovery rights under Rules 26 and 30” (source).

The people sanctioned: three attorneys, publicly reprimanded and disqualified. The order states the court “PUBLICLY REPRIMANDS Mr. Reeves, Mr. Cranford, and Mr. Lunsford for making false statements to the court; ORDERS publication of this order to effectuate that reprimand; DISQUALIFIES them from further participation in this case; and REFERS this matter to the Alabama State Bar and other applicable licensing authorities” (source).

The money: no monetary fine. The court reasoned that “a fine would not have rectified egregious misconduct in [the] present case” and imposed no monetary sanction (source).

The court declined to punish the firm: it “RELEASES WITHOUT SANCTION attorneys Daniel J. Chism and Lynette E. Potter, and the law firm Butler Snow LLP (‘Butler Snow’) from disciplinary proceedings” (source). An independent account notes the court credited Butler Snow’s pre-existing AI policies and its post-incident response in declining to sanction the firm (source). Manasco also ordered the three to disclose the sanctions order to their clients, opposing counsel, and the presiding judges in every other case in which they appear (source).

How this was verified

Every figure here was re-checked on 2026-08-17 against the court’s own published opinion, Johnson v. Dunn, 792 F. Supp. 3d 1241 (N.D. Ala. 2025), read in full from a hosted copy of the Thomson Reuters reprint: the five-citation count, the “completely made up” and “substantially greater accountability” language, the operative reprimand/disqualification/referral paragraph, the no-fine reasoning, and the release of Chism, Potter, and Butler Snow without sanction all appear verbatim in that order (Tier 1 primary). Two mutually independent secondaries corroborate the same facts: EDRM’s legal-industry analysis (Tier 2) and the Associated Press wire as carried by Alabama Daily News (Tier 2). No figure required the parties to confirm anything; this is an adjudicated public record. The one caveat: the post-order docket posture (any appeal or motion to amend, and the disposition of the bar and other-jurisdiction referrals) is not re-checked here and is not claimed on this page.

Related case files

The sanction Manasco called insufficient is the norm elsewhere: in Coomer v. Lindell, two MyPillow attorneys drew $3,000 fines each for a brief with dozens of defective citations, the “reprimands and modest fines” model this court rejected. The Lacey v. State Farm sanction sits at the money end of the same spectrum, $31,100 imposed by a special master, showing how differently courts price the identical failure. And Noland v. Land of the Free is the other case courts treat as precedent-setting: a first published appellate opinion on AI-fabricated citations, where the response was a $10,000 fine rather than disqualification.

Sources

Tiers: Tier 1 primary (the court’s own signed opinion or order); Tier 2 strong secondary (independent reputable press or legal-industry analysis naming the parties). All URLs checked live 2026-08-17.

1. U.S. District Court, N.D. Alabama, Johnson v. Dunn, No. 2:21-cv-1701-AMM, sanctions order (Doc. 204), 2025-07-23, published as 792 F. Supp. 3d 1241 (2025) (Tier 1, the court’s own signed opinion; read in full from a hosted Thomson Reuters reprint). https://www.law.berkeley.edu/wp-content/uploads/archive/2025/12/Johnson-v-Dunn.pdf
2. EDRM, “When AI Policies Fail: The AI Sanctions in Johnson v. Dunn and What They Mean for the Profession,” 2025-07-31 (Tier 2, independent legal-industry analysis naming the parties and quoting the order). https://edrm.net/2025/07/when-ai-policies-fail-the-ai-sanctions-in-johnson-v-dunn-and-what-they-mean-for-the-profession/
3. Alabama Daily News (Associated Press), “Judge sanctions lawyers defending Alabama’s prison system for using fake ChatGPT cases in filings,” 2025-07-25 (Tier 2, independent newsroom carrying the AP wire). https://aldailynews.com/judge-sanctions-lawyers-defending-alabamas-prison-system-for-using-fake-chatgpt-cases-in-filings/