By Edward J. McIntyre

The author of this article, Edward J. McIntyre, is a member of the Ethics Committee of the California Lawyers Association and the San Diego County Bar Association Legal Ethics Committee. The views expressed herein are his own. He focuses his practice exclusively on legal ethics and professional responsibility. 

Ethics articles are provided regularly by LACBA’s longstanding Professional Responsibility and Ethics Committee. The views and opinions expressed in this publication are solely of the authors.

---

Can the definition of insanity—"doing the same thing over and over and expecting different results”—be attributed to a whole profession?[1] One would think not, but then …

We all smirked when we heard the tale of the unfortunate New York lawyer who represented an individual plaintiff in a June 2023 personal injury lawsuit against Avianca Airlines. He filed in state court, but the airline removed the case to the Southern District of New York. One issue was the 1999 Montreal Convention, an international treaty governing airline liability for international airline travel, including personal injury.

Unfortunately, the lawyer’s small firm did not have federal research tools. But his teenage son came to the rescue. He told dad about ChatGPT, a new and “cool” research app that had been released to the public to much fanfare at the end of November 2022. The lawyer used it and filed his brief. But the court said it couldn’t find some cases he cited. So the lawyer went back to ChatGPT. The result? It doubled down and gave him additional fake citations. All told, more than two dozen fake cases; quotes; citations; non-existent opinions. 

At the Order to Show Cause hearing, the lawyer representing the by-then much-maligned lawyer argued that the extensive national and international public humiliation he had suffered was sufficient punishment. The court added a $5,000 sanction. Those of us who thought we knew better shook out heads.

“How could he have been so careless …so, so incompetent!”

And yet …

Since May 2023, in just the intervening two years and five months, there are, as of December 19, 2025, 707 judicial decisions identifying cases in which generative AI has produced hallucinated content—typically fake citations but also other forms of AI-generated fabricated material.[2] Of those 707 cases, the vast majority, 482, are in the United States where federal and state trial and appellate courts have issued decisions, often imposing sanctions, including referrals to disciplinary authorities because of fabricated cases, false quotations, phony citations, fabricated exhibits and misrepresented precedents.[3] The closest runner up to us is Canada, with merely 50 such decisions. California alone has contributed more than four dozen cases to the United States total.

Nor is this phenomenon limited to lawyers in solo practice or smaller firms and pro se litigants. A couple of major, international firms—no names here—have been victims of their own misuse of generative AI in court filings. Even the federal judiciary is not immune. In July 2025, a district judge in the District of New Jersey withdrew an opinion after counsel pointed out that the court misstated case outcomes and had five fake quotations attributed to opinions that did not contain them. That same day, a federal judge in the Southern District of Mississippi withdrew an order after a lawyer pointed out AI-generated errors: fabricated citations; non-existent parties to the case; and, false quotes. Might we ponder were trusted law clerks the drafters of those judicial opinions?

We don’t need Mr. Charlotin’s research and data collection to bring this issue to our attention. Many of these cases receive their own publicity—for example, in Law 360 or The American Lawyer. In addition, some courts and judges have now imposed standing orders, requiring lawyers and pro se litigants to identify their use of generative AI—for example, judges in the Northern District of California; the Northern District of Texas; The United States Court of International Trade; the Northern District of Illinois; the Southern District of Ohio; the Eastern District of Pennsylvania; the Eastern District of Missouri; the Northern District of California; the District of Hawaii.      

In addition to public embarrassment, judicial displeasure and angry clients, what are the ethical implications of this, now oft repeated, conduct? We can start at the beginning.

California Rule of Professional Conduct1.1, Comment [1] provides that the duties stated in our competence obligation include “the duty to keep abreast of the changes in law and its practice, including the benefits and risks associated with relevant technology” [italics added]. With so much publicity about hallucinogenic responses from generative AI platforms, can any of us now be unaware of the risks of using them for legal research without double-checking its purportedly “on-point” citations?

Moreover, Rule 1.1 prohibits reckless and grossly negligent performance of legal services. Can filing a court document without first determining that the cases cited actually exist, that the cases in fact contain the language quoted, that they stand for the propositions for which they are cited not be reckless? Not be grossly negligent? These questions, perhaps, answer themselves.

Rule 1.3 prohibits reckless or grossly negligent failure to act with reasonable diligence. Rule 1.3(b) defines “reasonable diligence” as acting “with commitment and dedication to the interests of the client and does not neglect or disregard or unduly delay a legal matter entrusted to the lawyer.” Prompt delivery of legal services, however, without the corresponding care that rule 1.1 requires does not satisfy rule 1.3. The failure to determine—before filing a document with a court—that cited cases exist or that they have the quotations used would underscore a reckless lack of diligence in representing that client.

Rule 3.3(a)(1) prohibits knowingly making false statements of fact or law to a tribunal and failure to correct a previous false statement. Without question, citing cases that do not exist and quotations that are not in the cases cited falls within the “false statements of law” the rule states.

Rule 3.3(a)(2) prohibits knowingly misquoting to a tribunal the language of a book, statute, decision or other authority. The same conduct would appear to fall squarely within the scope of rule 3.3(a)(2).

But rule 3.3(a)(1) and 3.3(a)(2) require that the lawyer’s conduct be “knowingly.” Would mere sloppiness in using generative AI save a lawyer from , at least, a misconduct charge? The rules and the facts on the ground appear to answer that question.

First, Rule 1.0.1(f) in defining “[k]nowledge,” “know,” or ‘knows,” provides that a lawyer’s knowledge may be inferred from the circumstances.

Second, if we consider the circumstances that should surround a lawyer’s decision to use generative AI, the inferences become apparent. We know of 292, and counting, well-publicized judicial decisions in the United States alone—with more than two dozen in California—identifying instances where generative AI produced hallucinated content, including false case citations and non-existent cases. We know that courts are issuing standing orders directed at the  use of generative AI in filings.[4] And we know more generally from periodic publicity that generative AI can and does produce hallucinated results, not just in legal documents, but across a broad spectrum.[5]

With this background, could a lawyer at present credibly argue that the lawyer did not “know” the lawyer’s court submission contained phony cases or other false citations? Could the lawyer claim that the lawyer had stayed current about the risks of using generative AI? Could the lawyer in that situation possibly explain away a failure even to read cases cited to a court—to see if the case existed, or to see if it really had the quotes the lawyer included? Would “willful blindness” be a cogent response to an OSC or a defense to a discipline complaint? Let’s hope it is not a defense any of us will have to assert.

Will the State Bar’s Office of Chief Trial Counsel investigate or prosecute lawyers for misuse of generative AI in court filings? We don’t know. But that it has yet to do so publicly does not mean that, in an appropriate case, it will not do so or that it has not already begun to do so. In fact, in at least one published California Court of Appeal opinion, the court ordered the clerk of the court to forward to the State Bar a copy of its opinion, which imposed $10,000 in sanctions on the lawyer for misusing generative AI in producing the lawyer’s briefs.[6]

For California lawyers, however, the State Bar is not the only discipline forum about which they should be concerned. Each California federal district has its own discipline program that allows judicial officers to report lawyer misconduct.[7] Such complaints have—thus far rarely—resulted in trials and discipline, including suspension from practice in that district. Given, however, the number of federal judges and courts with generative AI standing orders, it is becoming clear that judicial tolerance for misuse of generative AI is waning. Could discipline follow?

The benefits of generative AI are obvious.[8] It can increase efficiency. It can reduce costs for clients. It can be a lifesaver for smaller firms coping with insurmountable discovery demands. Generative AI’s benefits, however, do not trump our ethical obligations to courts and clients to ensure that we review court submissions to ensure that every case cited exists, that every quote is accurate, that every representation is an accurate statement of cited precedent, that the “facts” we argue are truly facts. Let us not follow in the footsteps of those whose actions suggest a belief that doing the same thing over and over will produce different results.