Maryland Fake Citation Case – Court Issues a Warning – Referral to Atty. Grievance Comm’n.

Fake citation or “hallucination” cases have become common.  See, e.g., Doug Austin, What’s Causing These AI Hallucinations and How to Fix It (Oct. 13, 2025)(reporting 772 cases in academic literature); Doug Austin, AI Hallucination Cases: A Compiled List (Jun. 9, 2025); Damien Charlotin, AI Hallucination Cases Database – Damien Charlotin (listing 507 cases)(visited Nov. 2, 2025).

Maryland now has a reported opinion on the issue and the responsible attorney was referred by the Court to the Attorney Grievance Commission. The Court wrote:

The issues raised by Father typically would not result in a reported opinion. We report this case, however, to address a problem that is recurring in courts around the country, i.e., the use of AI to draft briefs or other pleadings, resulting in incorrect, inaccurate, or fictitious case citations. Mother’s brief in this case is replete with citation irregularities. These irregularities include citations to multiple fictitious cases, as well as misquoted passages and citations to cases that do not support the proposition for which they are cited. Before addressing the merits of the issues raised, we address this problem. We do so as a warning to others and to determine the court’s appropriate response in this case.

Mezu v. Mezu, __ Md. App. __, 2025 WL 3022460, at *1 (Appl. Ct. Md. Oct. 29, 2025)(Graeff, J.)(emphasis added).

The Appellate Court reviewed the briefs in a family law case, found citation irregularities, and “this Court issued an Order to Mother’s counsel to Show Cause (‘OSC’) why he should not be sanctioned and/or referred to the Attorney Grievance Commission. The OSC ordered Mother’s counsel to provide the Court with a written submission in the form of a sworn declaration and provide a detailed explanation as to how Mother’s brief was generated, how counsel located the fictitious cases, and why counsel cited cases that did not stand for the proposition cited.” Id. at *1.

The mother’s attorney essentially tried, in my words, to shift responsibility to a law clerk:

In his response to the OSC, counsel stated that he prepared the brief in collaboration with a law clerk who had worked in his office since 2021. Counsel stated that he “does not use Lexis and/or Westlaw because he does very little appellate work.” He indicated that he relied primarily on treatises for research and then retrieved the cited cases from the internet. Mother’s counsel stated that he had “never used artificial intelligence for any professional purpose.” After reviewing the law clerk’s initial draft, counsel “inquired about her research process.” The law clerk advised that she had located cases on the internet, printed case notes, and verified the citations. Mother’s counsel stated that he “was not involved directly in the research of the offending citations,” and he believed that he complied with the Maryland Rules of Professional Conduct based on his law clerk’s “efforts at vetting the referenced cases.” He acknowledged, however, that “his oversight was insufficient to ensure that all citations were verifiable Maryland authority.”

Id. at *2.

The law clerk’s affidavit stated that:

First, she “utilized AI in her research, explaining that she began by using ChatGPT to search for relevant cases. She received a generated list of cases, which she acknowledged included cases that were inaccurately cited and did not exist.”

Second, “[s]he then ‘searched for the case opinions via a basic Google search in order to extract proper citations, on-point verbiage, and to identify further cases cited within the proffered case.’”

Third:

After reviewing those results, the law clerk initiated a free trial of a “for-pay” website called “VLex,”[1] where she “was able to find the case opinions from the list provided by ChatGPT.” (Footnote omitted). She stated that her “examination of these case opinions did not raise any ‘red flags.’ ” The law clerk primarily used ChatGPT and VLex, but she also visited Court Listener,[2] CaseMine,[3] and Justia.[4] She stated that she “was not aware that these sites were controlled by AI, or that the search results produced by these sites were the product of AI.” The law clerk then finalized her draft and forwarded it to counsel.

Fourth:

Counsel then made changes to the draft and asked the clerk to review the latest version to ensure that the cases were accurately cited. The law clerk relied on ChatGPT to “review the citations for accuracy.” The affidavit states that counsel asked her multiple times whether she checked the case citations, and the law clerk assured counsel that she had complied with his requests.

The law clerk stated that:

[S]he believed that AI tools “essentially operate as a data or information aggregator that allowed for more in-depth and extensive research results.” She only learned the risks of hallucinated AI results after counsel received the OSC. The law clerk stated that the topic of AI had never been raised either by her or Mother’s counsel. She never advised counsel that she had been using AI, and counsel never advised her to either use or avoid the use of AI.

Id. at *2.

At that time, the law clerk was not a Member of the Bar and the Court wrote that: “Under Maryland Rule 1-311(b), the signature of an attorney on a brief constitutes a certification that the attorney has read the brief, and ‘to the best of the attorney’s knowledge, information, and belief there is good ground to support it.’”  Id. at n. 3.

The Court wrote that:

Recently, courts across the country have been presented with briefs containing “fraudulent legal citations.” [citation omitted]. This case, however, appears to be the first where the Maryland appellate courts have addressed the problems that occur when lawyers use AI, without the diligence required, in drafting briefs. The failure to use AI responsibly in legal research raises ethical issues and can result in sanctions when used improperly. It is unquestionably improper for an attorney to submit a brief with fake cases generated by AI.

Id. at *3 (emphasis added).

The Court wrote that AI “may be a valuable tool.” Id.  However, it also noted the need for informed use and cited statistics on “hallucinations.”

Further, the Court wrote: “The citation of fake cases in a brief or other pleading filed with a court implicates multiple Maryland Rules.”  Id. (emphasis added).  Chief among those rules is Rule 1-311(b), the signature certification rule. The Court wrote: “Implicit in the requirement that an attorney certify that there is good ground to support the contentions raised is that the attorney has read the primary cases on which the attorney relies for the argument submitted.” Id. at *3 (emphasis added).

It wrote:

Here, Mother’s counsel admitted that he did not read the cases cited. Instead, he relied on his law clerk, a non-lawyer, who also clearly did not read the cases, which were fictitious.

Counsel’s conduct here implicates several of the Maryland Attorney’s Rules of Professional Conduct. Rule 19-303.1 mandates that attorneys bring or defend only meritorious issues. It provides that an attorney “shall not bring or defend a proceeding or assert or controvert an issue therein, unless there is a basis for doing so that is not frivolous.” A “citation to a fake opinion does not provide a non-frivolous ground for” bringing or defending a proceeding.

Id. at *4 (citation omitted).

Here, counsel admitted to not having a Lexis or Westlaw subscription because he “does very little appellate work.” Instead of adequately preparing himself regarding the law, counsel left this work, in part, to his law clerk, a non-lawyer, and then failed to verify the accuracy of the research. He stated that he “was not involved directly in the research of the offending citations,” yet he did not confirm how the cases that were cited were obtained, without Lexis or Westlaw, and he asked only if the law clerk had verified the citations to the cases, which he had not read. In our view, this does not satisfy the requirement of competent representation. A competent attorney reads the legal authority cited in court pleadings to make sure that they stand for the proposition for which they are cited.

Id.  Additionally, Rule 19-305.3(b) required that the attorney supervise the non-attorney.

The Court noted that the attorney had taken corrective action, however:

Nevertheless, it would be a stretch to say that he was remorseful for his failures and the extra work that he caused in this case. Indeed, when questioned at oral argument about the concern that he clearly did not read the cases before submitting his brief (because the fictitious cases did not exist), counsel did not appear to share that concern, stating that he typically did not read cases he cites, but instead, he relies on summaries found in treatises or on the internet.

Id. at * 4 (emphasis added).

The Appellate Court cited a California decision stating that “[s]imply stated, no brief, pleading, motion, or any other paper filed in any court should contain any citations—whether provided by generative AI or any other source—that the attorney responsible for submitting the pleading has not personally read and verified.” Id. (citation omitted).

The Court held that counsel here “falls short of” professional responsibilities, stating:

Submitting fake cases to the court in a legal brief or other pleading can result in multiple harms. An attorney’s failure to comply with the attorney’s ethical obligations undermines the integrity of the judicial system, and it can hurt the client’s case. Here, in addition to detracting from counsel’s credibility, addressing this issue required counsel to take time at oral argument that otherwise could have been devoted to the merits of Mother’s case. Counsel’s conduct here also required this Court to take time to try to find the fake cases cited in Mother’s brief, and then research how other courts have dealt with situations involving similar attorney misconduct, diverting judicial resources from other pressing work.

Id. at *5.

The Court exhaustively considered how other courts had handled similar situations. Id. at *5-6. After holding that it would (or could) not enter a monetary sanction, and that submittal of a corrected brief would be unfair to the opposing party, the Court applied corrective measures under Rule 18-102.15(a). It wrote:

In determining whether to refer counsel to the Attorney Grievance Commission, we have considered the nature and severity of the conduct in this case. The brief here did not contain an isolated instance of a citation mistake. Rather, it cited multiple cases that did not exist, as well as others that did not support the proposition for which they were cited. When asked at oral argument about the concern when a lawyer obviously has not read the cases cited to the court, counsel indicated that this was not a unique circumstance, stating that he typically did not read the cases he cited in pleadings submitted to the court. Based on all the circumstances, we shall refer this case to the Attorney Grievance Commission.

Id. at *6 (emphasis added).  The Court then addressed the appellate issues. Id. at *6, passim.

As what I would call a postscript, in note 6, the Court wrote: “Citing fake cases also may cause the opposing party to waste time and money addressing this issue. This does not appear to be the case here. Father did not address the issue in his reply brief and appears to have discovered it only after this Court issued the OSC.”[5]

Maryland has a growing body of law on AI:

• Artificial Intelligence in Maryland Courts (Aug. 24, 2025);
• Maryland’s Facial Recognition Technology Statute (Aug. 14, 2025);
• Criminal Conviction Reversed After State Failed to Timely & Fully Disclose its Use of a Type of Artificial Intelligence (Aug. 11, 2025).

One excellent source of information on AI is the Sedona Conference’s free publication.  See A Review of Sedona’s “Artificial Intelligence (AI) and the Practice of Law” by The Hon. Xavier Rodriguez (Sep. 27, 2023).

The Maryland State Bar Association “offers Decisis as a complimentary legal research member benefit ($1,740 value). Decisis shares the same parent company as LexisNexis® and offers an expansive library which includes both state and federal cases, statutes, court rules, and more. Decisis provides MSBA Members with an easy-to-use interface, a reliable citator, and 7 day a week live customer support.”  Legal Research | Maryland State Bar Association.  The website states that: “Decisis uncovers the citations that can stand up to the scrutiny of the judge. If it comes up on Decisis, you can have the confidence it’s good law.”

Maryland has not expressly adopted a “technological competence” rule. See 40th State Adopts a Duty of Technological Competence – Is It a Good Idea? (Mar. 31, 2022).  Rule 19-301.1 states that: “An attorney shall provide competent representation to a client. Competent representation requires the legal knowledge, skill, thoroughness and preparation reasonably necessary for the representation.”  Comment [6] states: “To maintain the requisite knowledge and skill, an attorney should keep abreast of changes in the law and its practice, engage in continuing study and education and comply with all continuing legal education requirements to which the attorney is subject.”

Doug Austin’s solution to the use of AI is training and education:  “We need to treat gen AI models and other forms of automation with a similar level of caution we apply to handling hazardous materials or operating heavy machinery.”  He adds:

What’s causing these AI hallucinations? Something inherent in all of us – the desire to reduce effort, especially when we need to get things done quickly. Would you skip steps when handling hazardous materials? No, because we’ve all been trained to be careful with them. We need to learn to be just as careful when working with gen AI models, especially public LLMs like ChatGPT that are a “jack of all trades, master of none” when it comes to things like accurate case citations. Handle with care!

Doug also wrote: “The problem is a phenomenon known as ‘automation bias’, which is the tendency for people to believe a result just because it came from a computer program or algorithm. One example of automation bias is ‘Death by GPS’ where people drive their cars into a lake or off an unfinished bridge, just because the GPS told them to go that way (or they put their Tesla in ‘autopilot’ mode and went to sleep). In fact, the National Highway Traffic Safety Administration (NHTSA) estimates that GPS causes over 200,000 car accidents every year in the US.”  What’s Causing These AI Hallucinations and How to Fix It.

____

[1] vLex | Legal AI

[2] CourtListener.com

[3] Casemine

[4] Justia

[5] It appears that a “bypass” petition for certiorari was dismissed. 492 Md. 396 (Aug. 22, 2025).