
Louisiana Attorney General Liz Murrill told a federal court Wednesday that her office will not comply with a judge's request that the state apologize to the Southeastern Conference over claims that SEC attorneys used “hallucinated” citations generated by artificial intelligence. Murrill filed a letter in Alabama's Northern District Court stating that forcing the state to publish a court-directed apology would cross a constitutional line.
In her filing, Murrill argued that the court's demand to publish a particular message, and submit it for judicial editing, violates basic free speech principles and does not satisfy strict scrutiny, according to AL.com. She asserted that government officials retain a free-speech right that includes both the right to speak and the right not to speak, and said her office simply could not comply with the court's directive.
U.S. District Judge Annemarie Axon had asked LSU and the Louisiana attorney general's office to issue public apologies retracting their claim that SEC attorneys relied on hallucinated citations, requesting what the outlet described as “heartfelt, robust” press releases essentially retracting the assertion. The dispute traces back to an SEC motion filed in October seeking to strike the hallucinated-citations assertion, with SEC lawyers calling the suggestion baseless, inflammatory, and wholly improper.
How the Word Hallucinated Became a Flashpoint
The fight centers on language LSU's attorneys used describing certain SEC citations. LSU's lawyers have since filed a motion to replace the phrase describing citations that “appear to be negligently supplied or hallucinated because they” with the word “actually,” a softened substitution that has not resolved the underlying dispute.
LSU attorney Bruce Rogers told the court the word hallucinated was standing for something broader than just a fake, AI-generated citation, the same AL.com report noted. But per ESPN, Axon clarified with Rogers that the citation in question was in fact legitimate, and Rogers responded, “They're not made-up cases.”
Axon did not accept Rogers' framing of the dispute as legitimate advocacy. She rejected his argument that the hallucinated-citations claim amounted to ordinary litigation strategy rather than an attempt to create a media firestorm and damage the professional reputations of SEC lawyers, according to the same report. The judge also told Rogers, in blunt terms, that he had put the pig out there and could not put lipstick on it.
SEC's Use of AI Tool Harvey Adds Another Layer
Part of what fueled the dispute is the SEC's own disclosure that it used the legal AI program Harvey to help create its brief. The conference certified that its attorneys independently checked the program's work before filing, the article notes, a detail Murrill cited in arguing that the hallucinated-citations claim constituted legitimate advocacy rather than defamation, since the SEC had acknowledged its use of artificial intelligence in the first place.
Murrill also asked Axon to relieve LSU's local counsel, Bruce Rogers, from his role in the case and to waive the local-counsel requirement altogether. Meanwhile, the stakes in the broader litigation remain significant: the SEC is seeking court approval to enforce its conference rules, a process that could potentially include expelling LSU from the conference.
Part of a Wider Legal Reckoning Over AI
The LSU-SEC fight is unfolding against a backdrop of mounting court battles over AI-generated errors in legal filings nationwide. A Norton Rose Fulbright update tracked six separate decisions concerning Gen AI hallucinations issued between February and April 2026 alone, and cited a researcher who has documented more than 1,148 cases of lawyer hallucinations tied to AI tools. A separate tracking project, Damien Charlotin's hallucinations database, tracks instances where a court or tribunal explicitly found or implied reliance on hallucinated material.
Other recent cases show how unevenly courts have responded to AI-related filing errors. Sullivan & Cromwell apologized to a federal judge after AI-generated errors, including fabricated citations, turned up in a court filing, providing a ledger of mistakes that spanned three pages and totaled around three dozen, according to The New York Times, which described the legal profession as undergoing a broader reckoning over the growing and widespread use of AI.
Consequences elsewhere have ranged from modest to severe. In Fletcher v. Experian Information Solutions, Inc., the Fifth Circuit imposed a $2,500 monetary sanction after finding fabricated quotations and other serious misrepresentations in a brief, per the Norton Rose Fulbright update. In Gamez v. County of Fresno, by contrast, a district court discharged a show-cause order without imposing sanctions after accepting an attorney's explanation that they believed the AI tool had anti-hallucination protection built in. An Eastern District of New York magistrate judge similarly declined monetary penalties against a lawyer who submitted three hallucinated cases, citing her remorse and personal circumstances, according to the ABA Journal. An Arizona case went the other direction entirely, with a magistrate judge revoking an attorney's pro hac vice status, removing her as counsel of record, striking her opening brief, and ordering apology letters to three judges over faulty citations.
Whether Axon will ultimately grant Murrill's request to remove Rogers from the case, or how she will respond to the state's refusal to apologize, remains to be seen as the underlying SEC enforcement dispute continues to play out in court.









