Lawyers can face suspension when court filings contain AI-generated fake citations, but a recent disciplinary order from the U.S. Court of Appeals for the Ninth Circuit makes clear that artificial intelligence itself is not the misconduct.
The court held that the violations arose when attorneys signed and filed briefs containing fabricated authorities, misattributed quotations and serious inaccuracies, rather than from the mere use of generative AI during drafting.
The June 2026 order provides one of the most detailed judicial examinations yet of how traditional professional responsibility rules apply to generative AI. It also sends a warning to law firms increasingly incorporating AI into legal research and drafting workflows: courts are unlikely to treat artificial intelligence as a defence when inaccurate legal authorities find their way into court filings.
The decision is significant because it answers a question confronting legal departments, law firms and regulators across the profession. As generative AI becomes more common in legal practice, when does an AI-assisted mistake become professional misconduct?
The Ninth Circuit’s Disciplinary Order
The Ninth Circuit suspended California attorneys Mike Singh Sethi and William Rounds from practice before the court for six months after determining that briefs filed in an immigration appeal contained nonexistent cases, misattributed quotations and significant misrepresentations of actual legal authorities.
The court also imposed personal sanctions of $2,500 against each lawyer, ordered disclosure of the disciplinary order to clients and courts in pending matters, and referred the matter to relevant licensing authorities.
The sanctions arose from Lnu v. Blanche, an immigration appeal in which the court found that the attorneys submitted authorities that either did not exist or did not support the legal propositions for which they were cited.
According to the order, the opening brief cited two nonexistent cases and attributed quotations to real decisions that contained no such language. Additional concerns later emerged in reply briefing and in filings submitted in other Ninth Circuit cases.
The panel found that the problems extended beyond the original filings. During oral argument, the attorneys initially characterised the errors as typographical or copy-and-paste mistakes and denied that generative AI had been used.
The court later concluded that those explanations were false and found that the attorneys had knowingly or recklessly made misrepresentations to the court, resulting in violations of their duty of candour toward the tribunal.
In response to the court’s order to show cause, the attorneys subsequently acknowledged that it was more likely than not that unauthorised AI use by unlicensed brief writers within the firm had contributed to the errors and that the citations had not been adequately verified before filing.
Why the Court Said AI Was Not the Violation
The Ninth Circuit began by addressing a point that sits at the centre of the debate over AI in legal practice. The judges made clear that they were not disciplining the attorneys simply because generative artificial intelligence had been used during the preparation of court filings.
Instead, the panel stressed that the same professional and procedural obligations apply regardless of whether a brief is drafted with the assistance of AI, a junior employee, another lawyer or traditional legal research methods. Responsibility ultimately rests with the lawyer whose signature appears on the filing.
The order repeatedly returns to that principle. According to the panel, professional rules are not breached when research is conducted or a draft is prepared. The problem arises when an attorney signs and submits a document containing authorities that do not exist, quotations that cannot be found in the cited case, or descriptions of legal precedent that bear little resemblance to what the underlying decision actually says.
Although AI generated many of the headlines surrounding the case, the judges focused on something far more familiar: whether lawyers fulfilled the professional obligations that already existed long before generative AI entered legal practice. That distinction runs throughout the order and provides the foundation for the disciplinary findings that followed.
Legal Framework and Potential Liability
Rather than creating new standards for AI-assisted legal work, the Ninth Circuit relied on existing professional conduct rules. Because both attorneys maintained their principal office in California, the panel applied the California Rules of Professional Conduct when evaluating the conduct at issue.
The judges concluded that the attorneys failed to meet their duties of competence and diligence after filing briefs that contained fabricated authorities, misattributed quotations and inaccurate descriptions of legal precedent. A recurring theme in the order is that lawyers must read and verify the authorities they cite before filing documents with a court. Delegating research or drafting does not transfer that responsibility to someone else.
The panel also rejected the suggestion that the fabricated citations were harmless because similar authorities could have been cited instead. A lawyer’s obligation is to cite the authorities actually relied upon. Cases that do not exist cannot support a legal argument, and identifying a comparable case after the problem is discovered does not cure what was originally filed with the court.
The most serious findings involved candour. After the filing errors came to light, the attorneys initially described the problems as typographical or copy-and-paste mistakes and repeatedly denied that generative AI may have been involved.
The panel concluded that those explanations were not credible and found that the attorneys had knowingly or recklessly made false statements to the court. The panel stated that the gravity of the discipline imposed, including the six-month suspension, was owed to the attorneys’ repeated failures of candour after the filing errors were discovered.
There is also a broader lesson for law firms. The order notes that substantial portions of the briefs were prepared by unlicensed brief writers and questions whether the firm’s review process provided adequate supervision.
While much of the attention surrounding the decision has focused on AI, the discussion of supervision and verification may prove equally significant for firms that rely on delegated drafting work, whether technology is involved or not.
When AI Errors Become Professional Misconduct
The case arrives at a time when many law firms are experimenting with generative AI for legal research, drafting and document review. While headlines have focused on fabricated citations, the Ninth Circuit’s analysis addresses a broader issue: how lawyers supervise work and verify legal authorities before documents are filed with a court.
A recurring theme throughout the order is that professional judgment cannot be delegated. The judges stressed that lawyers must read the authorities they cite, verify quotations and ensure that legal arguments are supported by the cases referenced in their filings. That responsibility does not disappear because research was performed by a colleague, an unlicensed brief writer or an AI tool.
The panel also highlighted a risk that extends beyond obviously fabricated cases. A nonexistent authority may be relatively easy to identify. A real case that is inaccurately described can be far harder to detect, particularly where the authority appears relevant at first glance but does not actually support the proposition for which it is cited.
According to the court, these inaccuracies often require close examination of the underlying authorities and may present a greater long-term challenge for legal practice than outright hallucinated cases.
The order offers insight into how at least one federal appellate court is applying existing professional responsibility rules to AI-related issues in legal practice. Rather than waiting for AI-specific legislation or new professional conduct rules, the Ninth Circuit applied longstanding duties of competence, diligence, candour and supervision.
The court’s message was that existing professional obligations remain fully applicable, regardless of the technology used during preparation of a filing.
Perhaps most significantly, the judges placed considerable weight on what happened after the errors were discovered. The panel suggested that prompt disclosure and transparency may have resulted in lesser sanctions. Instead, it concluded that repeated failures of candour justified much of the discipline ultimately imposed.
At the centre of the Ninth Circuit’s order is a principle that predates generative AI. Lawyers remain responsible for the accuracy of the documents they file with a court. Every citation, quotation and legal representation ultimately falls back on the attorney whose name appears on the filing, regardless of whether the document was prepared through traditional research, delegated drafting or AI-assisted workflows.