Amarsingh v Frontier Airlines sanctions for AI citations

Courts Impose Sanctions on Lawyers for AI-Generated Hallucinated Citations in Four High-Profile Cases

In late 2025, federal and state courts across the United States issued sanctions against lawyers in four distinct cases—Amarsingh v. Frontier Airlines, Lexos Media v. Overstock, Withers v. Blanche—due to the inclusion of fabricated or unverified legal citations generated by large language models (LLMs).

The cases span civil litigation, intellectual property disputes, and immigration proceedings, demonstrating that AI hallucinations in legal citations are not confined to a single practice area but have permeated multiple levels of U.S. federal and state courts. Judges responded with fines, referrals to disciplinary authorities, and standing orders requiring transparency in AI use, signaling a shift toward stricter enforcement of epistemic integrity in legal practice.

The sanctions underscore a critical failure: lawyers either relied on AI-generated content without verification or failed to exercise due diligence in reviewing filings they signed. Courts explicitly linked these errors to the use of tools like ChatGPT, First Drafts, and other AI-assisted legal research platforms, reinforcing that the responsibility for accuracy lies with legal professionals, not the technology itself.

Amarsingh v. Frontier Airlines: Appellate Brief Errors Lead to $1,000 Fine and Disciplinary Referral

Frontier Airlines, the Tenth Circuit Court of Appeals sanctioned attorney Kusmin L. Amarsingh for submitting an appellate brief containing fabricated case citations and misattributed legal authorities generated by ChatGPT. The court identified the infraction on January 24, 2025, after the brief was filed, and issued a show-cause order on September 5, 2025. The final ruling on February 9, 2026, affirmed the district court’s judgment, ordered Amarsingh to pay $1,000 to Frontier Airlines, and referred the matter to Maryland’s disciplinary authorities for further review.

The Tenth Circuit’s decision highlighted that Amarsingh had used ChatGPT to draft portions of the brief but failed to verify the citations before filing. The court emphasized that such reliance on unverified AI output undermines judicial integrity and the adversarial process, particularly when opposing counsel or the court cannot independently confirm the cited authorities.

The incident also revealed a pattern: lawyers using AI tools often assume the technology’s outputs are accurate without cross-referencing primary sources.

Lexos Media v. Overstock: Combined $12,000 Fine for Unverified AI-Generated Filings

In Lexos Media IP LLC v. Overstock.com Inc., a Kansas federal district court imposed a combined $12,000 fine on five lawyers after determining that their filings contained defective legal citations generated through AI tools without prior verification. The court denied corrections in December 2025 and formally sanctioned the lawyers on February 2, 2026.

The sanctions order noted that the AI-generated citations included nonexistent case names and misquoted legal precedents, which the opposing party and the court could not locate.

The court’s ruling in Lexos Media v. Overstock introduced a financial penalty structure tied to the severity of the errors, with the total fine distributed among the responsible lawyers.

Withers v. Aberdeen: Four Lawyers Sanctioned and Removed Over AI Hallucinations in Opposing Filings

City of Aberdeen presented a unique scenario where both sides of a civil litigation case were sanctioned for AI-related errors. A federal judge in Mississippi found that four lawyers—Kathleen M. Williams, Shauncey Hunter Ridgeway, and Mark McClinton—had submitted filings containing hallucinated legal citations. The court determined that Wilson used First Drafts to draft a filing, while Williams used an AI legal-research tool. Ridgeway and McClinton were sanctioned for failing to verify the filings they signed or authorized, despite knowing AI had been involved.

The sanctions order, issued on June 8, 2026, removed all four attorneys from the case and imposed fines. The ruling also noted that the errors had the potential to mislead the court and the opposing party, further degrading the reliability of legal proceedings.

This case introduced an additional layer of accountability: lawyers who sign or authorize filings drafted with AI assistance are equally responsible for verifying their content, regardless of whether they directly interacted with the AI tool. Aberdeen served as a warning to legal teams that AI integration must be accompanied by rigorous human review processes to prevent systemic errors.

LNU v. Blanche: Immigration Attorneys Sanctioned for Fabricated Citations and Lack of Candor

Blanche, the Ninth Circuit Court of Appeals sanctioned California immigration attorneys Mike Singh Sethi and William Rounds for submitting briefs containing fabricated legal authorities. The court concluded that the errors were likely generated by unauthorized AI tools and that the attorneys had failed to verify the citations or disclose the potential AI involvement when questioned. The disciplinary order, filed on June 3, 2026, cited repeated instances of the lawyers’ lack of candor regarding the source of the errors.

The Ninth Circuit’s ruling distinguished itself by focusing not only on the technical errors but also on the attorneys’ failure to correct the record or acknowledge the AI’s role when confronted. The court stated that the use of AI-assisted drafting systems without proper oversight or disclosure constitutes professional misconduct, particularly in cases involving detained clients where the stakes of legal accuracy are heightened. This case highlighted the intersection of AI risks and ethical obligations in immigration law, where errors can have direct consequences for individuals’ liberty and rights.

This case became a cautionary example for attorneys in sensitive practice areas where the reliability of legal arguments directly impacts clients’ legal status.

Judicial Responses and the Evolution of AI Disclosure Requirements

Across these four cases, courts adopted a consistent approach: sanctions were not merely punitive but were designed to enforce transparency and verification in AI-assisted legal work. These standing orders reflect a broader judicial trend toward mandating AI transparency, similar to existing rules governing conflicts of interest or ex parte communications.

Aberdeen, the removal of attorneys from the case demonstrated that judicial trust is not merely eroded by errors but by a failure to uphold basic professional standards.

The sanctions also revealed a pattern in the types of AI tools implicated: ChatGPT, First Drafts, and other large language models were consistently identified as the sources of hallucinated citations. Courts did not distinguish between different AI platforms but instead focused on the lawyers’ failure to implement verification protocols.

Broader Implications for Legal Practice and Epistemic Integrity

The 2025 sanction orders collectively signal a turning point in how courts view AI in legal practice. Two years after Mata v. Avianca established the precedent for AI-related errors, these cases demonstrate that judicial tolerance for unverified AI outputs has reached its limit. The harm identified in these incidents extends beyond individual cases: fabricated citations contribute to misinformation in legal arguments, erode public trust in the judiciary, and impose regulatory penalties on attorneys and firms.

For legal professionals, the takeaway is clear: AI tools must be treated as assistants, not substitutes, for legal research and drafting. The sanctions in these cases serve as a reminder that the integration of AI into legal workflows must be accompanied by robust verification processes to prevent errors that could have material consequences for clients and the judicial system.

Finally, these cases highlight the role of judicial discretion in shaping AI ethics in legal practice.

Sources

Incident summary from the AI Incident Database, used under CC BY-SA 4.0.

Contribute

Spot an error, or know a record we should add? Members can propose additions and submit corrections.

Submit incident Report a correction or request removal

Related in AI Incident Database

Sixth Circuit sanctions lawyers over fake appellate citations bearing the hallmarks of AI hallucination

In Whiting v. City of Athens the Sixth Circuit sanctioned counsel for briefs citing non-existent authority, one of the first appellate sanctions of the AI era.

Misinformation and fabricated claims2026

A US Department of Justice attorney files a brief containing AI-fabricated quotations and misstated holdings

A federal prosecutor's brief was found to contain quotations that do not exist in the cited cases, bringing the hallucinated-citation problem inside the Justice Department.