Rule 37 Sanctions Affirmed And Lawyer is Caught Citing Hallucinated Cases

Park v. Kim, 91 F.4th 610 was decided by the Second Circuit earlier this year. The district court dismissed plaintiff’s case under Rule 37 and Rule 41(b) because plaintiff violated court orders. On appeal plaintiff’s brief was found to contain hallucinated caselaw from artificial intelligence programs.

The dismissal was affirmed:

Over the course of the litigation before the District Court, Park continually and willfully failed to respond to and comply with the District Court’s discovery orders. Magistrate Judge Bloom issued numerous discovery orders and provided ample warning to Park and her counsel that timely compliance was expected. For example, at a telephonic conference on August 11, 2021, Judge Bloom granted Kim’s motion to compel, warning Attorney Lee: “[Y]our client can be subject to sanctions, which could be as severe as dismissal of the case, if she fails to comply.” J. App’x at 81-82.[1] On November 29, 2021, Judge Bloom again warned of the consequences of non-compliance, in no uncertain terms: “Plaintiff shall have one final opportunity to comply with the Court’s discovery Order…. This is a Court Order and plaintiff must comply. This is plaintiff’s last chance.” J. App’x at 7 (emphasis in original). In that same order, Judge Bloom set a briefing schedule for filing a motion to dismiss for non-compliance, should such a motion be necessary. Finally, having still not received the ordered discovery more than seven months after Judge Bloom’s August 2021 order, Kim moved to dismiss based on Park’s failure to comply with court orders and discovery obligations. See J. App’x at 292-93.

In her report and recommendation, Judge Bloom carefully considered all of the requirements of Rule 37 and Rule 41(b), including the availability of lesser sanctions, and concluded that dismissal was appropriate. Judge Chen, the presiding District Judge, reviewed Park’s objections to the report and recommendation in detail, overruled them, adopted the report and recommendation, and issued an order of dismissal on August 24, 2022.

The brief filed by plaintiff made things worse:

The reply brief cited only two court decisions. We were unable to locate the one cited as “Matter of Bourguignon v. Coordinated Behavioral Health Servs., Inc., 114 A.D.3d 947 (3d Dep’t 2014).” Appellant’s Reply Br. at 6. Accordingly, on November 20, 2023, we ordered Park to submit a copy of that decision to the Court by November 27, 2023. On November 29, 2023, Attorney Lee filed a Response with the Court explaining that she was “unable to furnish a copy of the decision.” Response to November 20, 2023, Order of the Court, at 1, Park v. Kim, No. 22-2057-cv (2d Cir. Nov. 29, 2023), ECF No. 172 (hereinafter, “Response”). Although Attorney Lee did not expressly indicate as much in her Response, the reason she could not provide a copy of the case is that it does not exist — and indeed, Attorney Lee refers to the case at one point as “this non-existent case.” Id. at 2.

Attorney Lee’s Response states:

I encountered difficulties in locating a relevant case to establish a minimum wage for an injured worker lacking prior year income records for compensation determination…. Believing that applying the minimum wage to in injured worker in such circumstances under workers’ compensation law was uncontroversial, I invested considerable time searching for a case to support this position but was unsuccessful.

Consequently, I utilized the ChatGPT service, to which I am a subscribed and paying member, for assistance in case identification. ChatGPT was previously provided reliable information, such as locating sources for finding an antic furniture key. The case mentioned above was suggested by ChatGPT, I wish to clarify that I did not cite any specific reasoning or decision from this case.

Id. at 1-2 (sic).

All counsel that appear before this Court are bound to exercise professional judgment and responsibility, and to comply with the Federal Rules of Civil Procedure. Among other obligations, Rule 11 provides that by presenting a submission to the court, an attorney “certifies that to the best of the person’s knowledge, information, and belief, formed after an inquiry reasonable under the circumstances… the claims, defenses, and other legal contentions are warranted by existing law or by a nonfrivolous argument for extending, modifying, or reversing existing law or for establishing new law.” Fed. R. Civ. P. 11(b)(2); see also N.Y. R. Pro. Conduct 3.3(a) (McKinney 2023) (“A lawyer shall not knowingly: (1) make a false statement of … law to a tribunal.”). “Rule 11 imposes a duty on attorneys to certify that they have conducted a reasonable inquiry and have determined that any papers filed with the court are well grounded in fact, [and] legally tenable.” Cooter & Gell v. Hartmarx Corp., 496 U.S. 384, 393, 110 S.Ct. 2447, 110 L.Ed.2d 359 (1990). “Under Rule 11, a court may sanction an 615*615attorney for, among other things, misrepresenting facts or making frivolous legal arguments.” Muhammad v. Walmart Stores E., L.P., 732 F.3d 104, 108 (2d Cir. 2013) (per curiam).

At the very least, the duties imposed by Rule 11 require that attorneys read, and thereby confirm the existence and validity of, the legal authorities on which they rely. Indeed, we can think of no other way to ensure that the arguments made based on those authorities are “warranted by existing law,” Fed. R. Civ. P. 11(b)(2), or otherwise “legally tenable.” Cooter & Gell, 496 U.S. at 393, 110 S.Ct. 2447. As a District Judge of this Circuit recently held when presented with non-existent precedent generated by ChatGPT: “A fake opinion is not `existing law’ and citation to a fake opinion does not provide a non-frivolous ground for extending, modifying, or reversing existing law, or for establishing new law. An attempt to persuade a court or oppose an adversary by relying on fake opinions is an abuse of the adversary system.” Mata v. Avianca, Inc., No. 22CV01461(PKC), 678 F.Supp.3d 443, 460-61 (S.D.N.Y. June 22, 2023).

Attorney Lee states that “it is important to recognize that ChatGPT represents a significant technological advancement,” and argues that “[i]t would be prudent for the court to advise legal professionals to exercise caution when utilizing this new technology.” Response at 2. Indeed, several courts have recently proposed or enacted local rules or orders specifically addressing the use of artificial intelligence tools before the court.[3] But such a rule is not necessary to inform a licensed attorney, who is a member of the bar of this Court, that she must ensure that her submissions to the Court are accurate.

Attorney Lee’s submission of a brief relying on non-existent authority reveals that she failed to determine that the argument she made was “legally tenable.” Cooter & Gell, 496 U.S. at 393, 110 S.Ct. 2447. The brief presents a false statement of law to this Court, and it appears that Attorney Lee made no inquiry, much less the reasonable inquiry required by Rule 11 and long-standing precedent, into the validity of the arguments she presented. We 616*616 therefore REFER Attorney Lee to the Court’s Grievance Panel pursuant to Local Rule 46.2 for further investigation, and for consideration of a referral to the Committee on Admissions and Grievances. See 2d Cir. R. 46.2.

We further ORDER Attorney Lee to provide a copy of this ruling to Plaintiff-Appellant Park — translated into Korean if necessary to permit Park to understand it — within twenty-one days, and to file a certification on the docket in this case attesting that she has done so.

Deleting Text Messages Leads to Rule 37 Sanctions And Dismissal

In Pable v. Chicago Transit Authority, 145 F.3d 712, the Seventh Circuit affirmed a grant of Rule 37 sanctions against a plaintiff who deleted text messages from his phone. Before 2015, I had never dealt with an issue involving the imaging of a client’s phone. Since then it has occurred in numerous cases that I have been involved with. My practice is to retain a third party to image the client’s phone.

Pable was employed by the CTA. After he was fired, he brought a whistleblower lawsuit against the CTA, under the National Transit Systems Security Act, 6 U.S.C. §1142. During the discovery process, the district court awarded Rule 37 sanctions, including dismissal of the case under Rule 37(e). Pable’s attorney was also sanctioned. The Seventh Circuit affirmed the decision.

The procedural background:

Pable later sued the CTA and Clever Devices (together “the defendants”) under the National Transit Systems Security Act, 6 U.S.C. § 1142, alleging that he was terminated 718*718 in retaliation for reporting a security vulnerability. The CTA responded by filing a counterclaim under the Computer Fraud and Abuse Act, alleging that Pable had configured a “doomsday” feature on his work computer that allowed him to wipe it remotely.[1] From there, the parties quickly descended into a protracted discovery battle that produced limited, but notable, evidence.

Most importantly, the parties learned that Pable and Haynes had used Signal to communicate about “what to do next” after the Dayton test. Pable testified that he used the application because it was more secure than text messaging and saved storage space on his phone. The record is silent on whether the CTA knew that Pable and Haynes were using Signal to communicate about their work or whether the CTA approved the use of Signal generally.

Pable’s explanations for the deletion of his Signal messages on November 2 evolved over time. Initially, Pable testified that the pre-November 2 messages vanished because Haynes deleted them from Haynes’s device. The CTA disputed this, providing an affidavit from Signal’s Chief Operating Officer stating that, at that time, one user’s deletion of specific messages did not remove those same messages from another user’s device. Confronted with this affidavit and a request for sanctions, Pable filed his own affidavit to explain that he had specially configured the Signal application on his device to delete threads when other users, like Haynes, deleted the entire thread. He added that he enabled that functionality for security reasons. The CTA, noting that Pable had not mentioned this explanation during his depositions, viewed this as a later-developed justification unworthy of credence.

The Signal saga did not end there. The CTA also learned during discovery that another group of Signal messages were lost. Specifically, Pable and Haynes had continued to communicate via Signal after their November 2, 2018, meeting. But nearly a year later, on October 29, 2019, Pable activated the “disappearing messages” feature in his Signal thread with Haynes. This caused the messages in that conversation to automatically delete twenty-four hours after being read by the recipient.

As the Signal issues unfolded, the CTA requested a forensic image of Pable’s phone to recover, among other information, the missing Signal messages. Pable’s counsel, Timothy Duffy, initially resisted, citing privacy concerns, but eventually agreed after extensive negotiation. On October 31, 2020, Pable produced the first image of his phone. The CTA’s expert, Nathan Binder, determined the image contained only 0.2 GB of user data, lacked messages from third-party applications, and omitted photos, browsing history, and other key categories of information. Duffy initially insisted that the image was complete but later acknowledged that the vendor might have imaged only relevant portions based on search terms. In a fight over whether a second imaging should be conducted, Duffy stated that he had instructed a third-party vendor, Quest Consultants International, to collect data based on agreed search terms and believed that the image was complete in terms of producing relevant information.

719*719 Duffy’s representations unraveled quickly. First, Dan Jerger, a Quest employee, testified that Duffy’s imaging instructions to Quest were limited to certain search terms and date ranges—rather than a full forensic image. Second, after another image was ordered, the CTA’s vendor recovered 25 GB of data—much more than the 0.2 GB earlier produced by Duffy. Remarkably, the CTA noted that the second image included discussions about the Skeleton Key vulnerability and other issues relevant to Pable’s suit. The second image, however, did not include any post-October 29, 2019, Signal messages between Pable and Haynes.

The limited information recovered from the second imaging was discovered through much effort, and the defendants moved for dismissal of the complaint and sanctions accordingly. They argued that Duffy unreasonably multiplied the proceedings through misrepresentations about the completeness of productions and failures to correct the record. They also argued that Pable failed to take reasonable steps to preserve three categories of Electronically Stored Information (ESI), including: (1) Signal messages exchanged with Haynes before November 2, 2018; (2) Signal messages exchanged after October 29, 2019; and (3) data on Pable’s personal cell phone.

The magistrate judge entered a report and recommendation finding in the defendants’ favor on those issues. The district court adopted the recommendation, dismissed Pable’s complaint with prejudice, and imposed a series of monetary sanctions. First, the court awarded the CTA sanctions pursuant to Federal Rule of Civil Procedure 37(e) for Pable and Duffy’s failure to preserve the relevant ESI. The penalty was $75,175.42, equally split between Pable and Duffy. Second, the court ordered Duffy to pay an additional $53,388 under 28 U.S.C. § 1927 for unreasonably and vexatiously multiplying the proceedings. Third, the court ordered Duffy to pay another $21,367 under Rule 37(a)(5) to compensate the CTA for having to file a motion to compel.

The Seventh Circuit held that the sanction of dismissal was appropriate:

When asked by the defendants to dismiss Pable’s action, the magistrate and district judges undertook reasoned analyses. The district court ultimately found that Pable intentionally spoliated two categories of ESI and dismissal was appropriate to remedy the resulting prejudice. There is, to be sure, some uncertainty about exactly how the messages before November 2, 2018, were deleted from Pable’s phone. And, on appeal, Pable seizes upon that uncertainty, arguing that the district court clearly erred in finding that he intentionally spoliated the ESI.

In addressing that argument, we look to the record. Haynes and Pable initially testified that Haynes alone deleted the conversation thread, and Pable testified that he believed that the deletion of the thread on Haynes’ phone caused the messages to be deleted from his phone as well. Although Pable’s explanation is at least plausibly consistent with the affidavit from Signal’s Chief Operating Officer (stating that Signal did not provide the ability for a single user to unilaterally and permanently delete specific messages for all users in a thread), the district court was within its discretion to not credit that explanation. After all, Pable’s explanation changed over time (first claiming that Haynes deleted the messages unilaterally, then adding that Pable had specifically configured the Signal application on his device to make a unilateral deletion possible). This evolution is notable because Pable made no mention of a custom configuration on his own phone when he offered his initial explanation, even though he would have known of that at the time and it would have been relevant to the deposition questions. Further, Pable offered no evidence other than his affidavit that such a configuration of Signal existed or was installed on his phone. Moreover, Pable’s explanation arrived only after he read the affidavit from the Signal COO refuting his prior explanation.

The district court also reasonably found that Pable’s duty to preserve those pre-November 2 messages was triggered prior to their deletion. Pable had reached out to attorneys about his termination days earlier on October 29, 2018. Then, in a follow-up email to one attorney the next day, Pable showed awareness that the CTA’s rationale for terminating him was relevant, writing “I think they are trying to find any reason to terminate me so I cannot take my FMLA and prevent payment of benefits.” And since he and Haynes were being interviewed at the same time, he must have realized that his actions regarding the Dayton test were at issue. These events, a factfinder could reasonably conclude, should have put him on notice that he needed to preserve the messages. Despite that notice, a factfinder could reasonably conclude, Pable deleted (or knowingly allowed Haynes to unilaterally delete) the messages. Although a different factfinder might have reached a different conclusion, the district court did not clearly err[2] in finding that Pable intentionally destroyed the ESI at issue.

721*721 Pable makes several other arguments about the dismissal sanction. First, he argues that he had a right to an evidentiary hearing because the district court had to resolve disputed facts and assess Pable’s credibility before dismissing his complaint. We review a district court’s decision not to hold a hearing for abuse of discretion. See REXA, Inc., 42 F.4th at 672. In this case, we see none because the court’s decision was based on its evaluation of undisputed, objective evidence: Pable allowed the messages to be deleted, his explanations for how exactly the messages were deleted evolved over time, and he took no reasonable steps to preserve the messages. The court was well within its discretion to conclude that an evidentiary hearing would not have assisted it in reaching a decision. See Kapco Mfg. Co. v. C & O Enters., Inc., 886 F.2d 1485, 1495 (7th Cir. 1989) (“[T]he right to a hearing in these circumstances is obviously limited to cases where a hearing could assist the court in its decision.”); see also REXA, Inc., 42 F.4th at 672-73 (“[A] court does not abuse its discretion by not conducting `an evidentiary hearing that would only address arguments and materials already presented to the court in the parties’ briefings.'” (quoting Royce v. Michael R. Needle P.C., 950 F.3d 481, 487 (7th Cir. 2020))). Moreover, Pable was given the opportunity to provide briefing on any of the arguments that he would have made at a hearing. For this reason too, we cannot say that the district court abused its discretion in imposing the sanction of dismissal without first holding an evidentiary hearing or submitting the credibility question to a jury.[3]

Skipping a hearing, Pable argues, was also inconsistent with our decisions in Kapco and McIntosh v. Wexford Health Sources, 987 F.3d 662 (7th Cir. 2021). But neither McIntosh nor Kapco do the work Pable suggests they do. Our court ruled against the Kapco appellants on the very issue for which Pable cites to it for support. There, we determined that the sanctions were proper and that due process did not require a hearing where the sanctioned attorney was given notice and an opportunity to respond before sanctions were imposed. Kapco, 886 F.2d at 1495. And McIntosh is entirely inapposite. Our holding in McIntosh—remanding with instructions to the district court to hold an evidentiary hearing before rejecting the magistrate judge’s credibility findings— turned on our reading of the Federal Magistrate Act and its constitutional implications. 987 F.3d at 664-66McIntosh is limited to that context, and should not be read to imply that litigants subject to sanctions must be afforded an evidentiary hearing before a district court can make credibility findings related to those sanctions.

The Seventh Circuit also affirmed 28 U.S.C. §1927 sanctions against Pable’s attorney.

Duffy argues on appeal that § 1927 sanctions were inappropriate because he did not unreasonably and vexatiously multiply the proceedings. He challenges the district court’s finding that he made misrepresentations, including by stating that the first image was a “complete image” and by later failing to correct that statement during the dispute over the need for a second imaging.

But the record before us supports the conclusion that Duffy’s statements were misrepresentations. To start, he affirmatively represented that the first image was a “complete forensic image,” as reflected in an email exchange between Duffy and counsel for the CTA. Indeed, he concedes as much on appeal, noting that, in meet-and-confer discussions and correspondence, “Duffy first represented the image was `complete’ on October 24, 2020.” He claims that his answer is taken out of context, but the context is clear as shown by the entire email exchange:

[CTA:] Pable will produce a complete and searchable forensic image file of his personal cell phone as it was previously imaged by Pable’s third-party expert during the course of written discovery in this litigation; as Pable has already had the phone imaged, we would ask that the image be produced by the close of business on October 29, 2020.

[Duffy:] The image is a complete forensic image; I cannot make any representations about its searchability, which has nothing to do with the imaging process. 726*726 I am checking on timing and logistics for this, and am just not in a position to confirm the 29th right now.

Duffy also argues that his statements are misrepresentations only if one “literally” interprets them. For example, he agrees that Jerger was not instructed to obtain a “complete image”—because, he continues, Jerger “was never asked to obtain anyimage.” This misses the point. Duffy agreed to image the phone and, on June 12, 2020, told counsel for the CTA that he had “imaged Mr. [Pable’s] cell phone and [was] in the process of running the search terms.” This was a misrepresentation by Duffy’s own admission because Duffy did not image the phone and knew that at the time, as Jerger testified and as Duffy acknowledges on appeal. Further, Duffy did indeed fail to correct his misrepresentations. There is no dispute on this point. Rather than correct the misrepresentations and resulting misunderstandings, he opposed each of the CTA’s efforts to produce a complete forensic image. Under these circumstances, Duffy cannot surmount the high hurdle of demonstrating that the district court abused its discretion in imposing sanctions under § 1927.

Comment: once the lawyer discovered that the client might have deleted messages on his phone he would have been wise to make no representations to the court and to disclose whatever instructions he gave to the data retrieval company. Failing that, there are many attorneys who know about these issues and who could have helped him before he became entombed in this mess. In my experience, search terms are negotiated in advance with the opposing party’s lawyers to avoid problems such as the ones that emerged here.

Ed Clinton, Jr.

http://www.clintonlaw.net

Bankruptcy Judge Finds AI citations Violate Rule 9011

Bankruptcy rule 11 is known as Rule 9011. In re Martin, 670 B.R. 636 (2025), issued by the Northern District of Illinois sanctions two lawyers for using artificial intelligence to prepare portions of their briefs. One lawyer claimed he was unaware of the prohibition on using artificial intelligence and that he did not know that it might make up fake citations to cases. The court’s response to this is thoughtful:

[Lawyer 1] and [Lawyer 2] ask me not to sanction them at all given that they have already voluntarily: (1) admitted their misconduct and promised not to do it again; (2) withdrawn any application for compensation in this case; and (3) watched an online CLE video. But while I appreciate their candor and efforts, “[t]here must be consequences.” Ferris v. Amazon.com Servs., LLC, No. 24-cv-304, 778 F.Supp.3d 879, 882 (N.D. Miss. Apr. 16, 2025). While I believe this mistake was unintentional, a “citation to fake, AI-generated sources … shatters [] credibility” and “imposes many harms.” Kohls v. Ellison, No. 24-cv-3754, 2025 WL 66514, at *4-5 (D. Minn. Jan. 10, 2025). So the consequences “are steep.” Id. at *5.

The first reason I issue sanctions stems from [Lawyer 1’s] claim of ignorance—he asserts he didn’t know the use of AI in general and ChatGPT in particular could result in citations to fake cases. (Dkt. No. 71 at 3) [Lawyer 1] disputes the court’s statement in Wadsworth that it is “well-known in the legal community that AI resources generate fake cases.” Wadsworth v. Walmart Inc., 348 F.R.D. 489, 497 (D.Wyo. 2025). Indeed, [Lawyer 1] aggressively chides that assertion, positing that 647*647 “in making that statement, the Wadsworth court cited no study, law school journal article, survey of attorneys, or any source to support this blanket conclusion.” (Dkt. No. 71 at 3-4, emphasis in [Lawyer 1’s] brief as filed)

I find [Lawyer 1’s] position troubling. At this point, to be blunt, any lawyer unaware that using generative AI platforms to do legal research is playing with fire is living in a cloud. This has been a hot topic in the legal profession since at least 2023, exemplified by the fact that Chief Justice John G. Roberts, Jr. devoted his 2023 annual Year-End Report on the Federal Judiciary (in which he “speak[s] to a major issue relevant to the whole federal court system,” Report at 2) to the risks of using AI in the legal profession, including hallucinated case citations.[6] To put it mildly, “[t]he use of non-existent case citations and fake legal authority generated by artificial intelligence programs has been the topic of many published legal opinions and scholarly articles as of late.”[7] At this point there are many published cases on the issue—while only a sampling are cited in this opinion, all but one were issued before June 2, 2025, when [Lawyer 1] filed the offending reply. See, e.g., Jaclyn Diaz, A Recent High-Profile Case of AI Hallucination Serves as a Stark Warning,NPR ILLINOIS (July 10, 2025, 12:49 PM), https://www.nprillinois.org/XXXX-XX-XX/a-recent-high-profile-case-of-ai-hallucination-serves-as-a-stark-warning (“There have been a host of high-profile cases where the use of generative AI has gone wrong for lawyers and others filing legal cases…. It has become a familiar trend in courtrooms across the U.S.”). The Sedona Conference wrote on the topic in 2023.[8] Newspapers, magazines, and other well-known online sources have been publicizing the problem for at least two years.[9] And on January 1, 648*648 2025, the Illinois Supreme Court issued a “Supreme Court Policy on Artificial Intelligence” requiring practitioners in this state to “thoroughly review” any content generated by AI.[10]

Counsel’s professed ignorance of the dangers of using ChatGPT for legal research without checking the results is in some sense irrelevant. Lawyers have ethical obligations not only to review whatever cases they cite (regardless of where they pulled them from), but to understand developments in technology germane to their practice.[11] And there are plenty of opportunities to learn—indeed, the Illinois State Bar Association chose “Generative Artificial Intelligence — Fact or Fiction” as the theme of its biennial two-day Allerton Conference earlier this year, calling the topic “one that every legal professional should have on their radar.”[12] Similar CLE opportunities have been offered across the nation for at least the past two years.

The bottom line is this: at this point, no lawyer should be using ChatGPT or any other generative AI product to perform research without verifying the results. Period. See, e.g., Lacey v. State Farm Gen. Ins. Co., No. CV 24-5205, 2025 WL 1363069, at *3 (C.D. Cal. May 5, 2025) (“Even with recent advances, no reasonably competent attorney should out-source research and writing to this technology— particularly without any attempt to verify the accuracy of that material.”); Mid Cent. Operating Eng’rs, 2025 WL 574234, at *2 (“It is one thing to use AI to assist with initial research, and even non-legal AI programs may provide a helpful 30,000-foot view. It is an entirely different thing, however, to rely on the output of a generative AI program without verifying the current treatment or validity—or, indeed, the very existence—of the case presented.”). In fact, given the nature of generative AI tools, I seriously doubt their utility to assist in performing accurate research (for 649*649 now). “Generative” AI, unlike the older “predictive” AI, is “a machine-learning model that is trained to create new data, rather than making a prediction about a specific dataset. A generative AI system is one that learns to generate more objects that look like the data it was trained on.” Adam Zewe, Explained: Generative AI, MIT NEWS (Nov. 9, 2023), https://news.mit.edu/2023/explained-generative-ai-1109 (emphasis added). Platforms like ChatGPT are powered by “large language models” that teach the platform to create realistic-looking output. They can write a story that reads like it was written by Stephen King (but wasn’t) or pen a song that sounds like it was written by Taylor Swift (but wasn’t). But they can’t do your legal research for you. ChatGPT does not access legal databases like Westlaw or Lexis, draft and input a query, review and analyze each of the results, determine which results are on point, and then compose an accurate, Bluebook-conforming citation to the right cases—all of which it would have to do to be a useful research assistant. Instead, these AI platforms look at legal briefs in their training model and then create output that looks like a legal brief by “placing one most-likely word after another” consistent with the prompt it received. Brian Barrett, “You Can’t Lick a Badger Twice”: Google Failures Highlight a Fundamental AI Flaw, WIRED (Apr. 23, 2025, 7:44 PM), https://www.wired.com/story/google-ai-overviews-meaning/.

If anything, [Lawyer 1’s] alleged lack of knowledge of ChatGPT’s shortcomings leads me to do what courts have been doing with increasing frequency: announce loudly and clearly (so that everyone hears and understands) that lawyers blindly relying on generative AI and citing fake cases are violating Bankruptcy Rule 9011 and will be sanctioned. [Lawyer 1] “professed ignorance of the propensity of the AI tools he was using to `hallucinate’ citations is evidence that [the] lesser sanctions [imposed in prior cases] have been insufficient to deter the conduct.” Mid Cent. Operating Eng’rs, 2025 WL 574234, at *3.

The second reason I issue sanctions is that, as described above, I also have concerns about the way this particular case was handled. I understand that Debtor’s counsel has a massive docket of cases. But every debtor deserves care and attention. Chapter 13 cases can be challenging to file and manage—especially when they involve complexities like those in this case. If a law firm does not have the resources to devote the time and energy necessary to shepherd hundreds of Chapter 13 cases at the same time, it should refer matters it cannot handle to other attorneys who can—lest a search for time-saving devices lead to these kinds of missteps. What I mean to convey here is that while everyone makes mistakes, I expect— as I think all judges do—attorneys to be more diligent and careful than has been shown here.[13]

Comment: This is an excellent opinion and it should be carefully read and considered by all lawyers.

Ed Clinton, Jr.

Even Pro Se Litigants Are Using Artificial Intelligence And Citing Hallucinated Cases

In Kent Malone-Bey v. Lauderdale County School Board, 3:25-cv-380, the plaintiff was terminated an instructor by the defendant. He filed suit. At some point he filed a motion to disqualify the law firm defending the case. The motion was denied. The opinion offers another chapter in the AI hallucination saga.

The Court sua sponte addresses a separate issue: Plaintiff’s continued use of fake, nonexistent legal opinions in his filings before the Court. This practice unfortunately is not unique to the instant matter as the number of filings that include fictitious legal opinions is rising. This trend, perhaps, correlates to the frequency at which our society is embracing and utilizing generative artificial intelligence (“AI”) programs in everyday life, including legal matters. AI is certainly a tool capable of providing immense benefits and efficiency in the legal profession or in legal matters. Those benefits, however, are not without risk.

Starting with the obvious, it is well-known that AI can generate fake sources of information, commonly referred to as “AI Hallucinations.” Wadsworth v. Walmart, Inc., 348 F.R.D. 489, 493 (D. Wy. 2025). And individuals who use AI for legal research find that they are not immune from these hallucinations. Indeed, AI programs are known to hallucinate nonexistent cases. Sanders v. United States, 176 Fed. Cl. 163, 169 (2025). As AI programs advance, they only increase the difficulty in which users (and the courts) may determine whether a case provided by an AI program is in fact “hallucinated.”

That is because hallucinated cases look like real cases. They are identified by a case name, a citation to a reporter, the name of a district or appellate court, and the year of the decision. United States v. Hayes, 763 F. Supp. 3d 1054, 1065 (E.D. Cal. 2025). But, they are not real cases. These hallucinated cases are instead inaccurate depictions of information from AI models that suffer from incomplete, biased, or otherwise flawed training data. Wadsworth, 348 F.R.D. at 493. Thus, “[w]hen used carelessly, [AI] produces frustratingly realistic legal fiction that takes inordinately longer to respond to than to create. While one party can create a fake legal brief at the click of a button, the opposing party and court must parse through the case names, citations, and points of law to determine which parts, if any, are true.” Ferris v. Amazon.com Servs., LLC, 2025 WL 1122235, at *1 (N.D. Miss. Apr. 16, 2025).

Here, Plaintiff’s Brief [18] overflows with either citations to fake, nonexistent cases or existing cases with incorrect legal propositions, i.e., the hallmarks of AI hallucinations. It therefore appears that Plaintiff may have used generative AI to assist in legal research and drafting his motion briefs.

As just one example, Plaintiff cites Coleman v. Ret. Plan for Dist. Managers, 969 F.3d 142, 149 (5th Cir. 2020) for the proposition that courts may disqualify counsel to “preserve the integrity of the adversary process.” [18] at 2. But the Coleman case does not exist.[5] By citing (and quoting) a fake opinion,[6] Plaintiff undermines the very integrity that he urges this Court to uphold. Indeed, submitting fictitious cases and quotations to the Court degrades or impugns the integrity of the Court. Hayes, 763 F. Supp. 3d at 1054. Moreover, Plaintiff’s attempt to persuade the Court “by relying on [a] fake opinion[] is an abuse of the adversary system.” Mata v. Avianca, Inc., 678 F. Supp. 3d 443, 461 (S.D.N.Y. 2023) (emphasis added).[7]

Even if Plaintiff did not use AI to assist with citing legal opinions and holdings, he still must comply with Federal Rule of Civil Procedure 11. See Yazdchi v. Am Honda Fin. Corp., 217 F. App’x 299, 304 (5th Cir. 2007). Rule 11 provides that “[b]y presenting to the court a pleading, written motion, or other paper — whether by signing, filing, submitting, or later advocating it — an attorney or unrepresented party certifies that to the best of the person’s knowledge, information, and belief, formed after an inquiry reasonable under the circumstances… the claims, defenses, and other legal contentions are warranted by existing law.” Fed. R. Civ. P. 11(b)(2).

By citing to fake legal opinions (or hallucinated legal holdings therein), Plaintiff not only wasted the Court’s time and judicial resources, but he also violated his Rule 11 obligations. See Benjamin v. Costco Wholesale Corp., 2025 WL 1195925, at *2 (E.D.N.Y. Apr. 24, 2025) (“an attorney [or pro se party] who submits fake cases clearly has not read those nonexistent cases, which is a violation of Rule 11 of the Federal Rules of Civil Procedure”) (emphasis in original). Plaintiff is therefore reminded of his obligations under Rule 11 as stated herein and advised that future filings before this Court must contain only accurate representations and citations. Additionally, Plaintiff is warned that future filings with citations to fictitious legal opinions and holdings or which do not otherwise comply with Rule 11 may result in sanctions, including but not limited to the striking of filings and the imposition of monetary penalties. Sanders, 176 Fed. Cl. at 170 (2025)Ferris, 2025 WL 1122235, at *2-3.

IT IS, THEREFORE, ORDERED that:

1. Plaintiff’s Motion to Disqualify Butler Snow LLP from Representing Individual Capacity Defendants [17] is DENIED; and

2. Plaintiff is reminded of his Rule 11 obligations as stated herein and warned that future filings with citations to fictitious legal opinions and holdings in filings before the Court, or which do not otherwise comply with Rule 11, may result in sanctions, including but not limited to striking Plaintiff’s filings and monetary penalties.

SO ORDERED.

Plaintiff’s Amendment of Complaint After Removal was insufficient to defeat diversity jurisdiction.

Plaintiff’s Amendment of Complaint After Removal was insufficient to defeat diversity jurisdiction.

Diversity jurisdiction requires that the plaintiff and defendant be citizens of different states and that the amount in controversy exceeds $75,000. Here the plaintiff sued State Farm for $100,000. State Farm removed the case. Plaintiff moved to remand it. The court denied the motion.

Plaintiff does not dispute that his initial pleading, which sought compensatory and punitive damages in excess of $10,000.00, as well as attorney’s fees, interest, and costs, satisfied the amount-in-controversy requirement for § 1332 diversity jurisdiction. See Compl. (Doc. No. 1-1) at 4; Notice of Removal (Doc. No. 1) at 10-12. Plaintiff argues that this case should be remanded to the state court, however, because his amended pleading establishes an amount in controversy that is explicitly “less than $75,000.00.” Am. Compl. at 2; see Pl.’s Mot. at 1-2. Defendant State Farm Auto responds that it is the demand in the initial pleading, not that of any subsequent amendment, that forms the appropriate basis for determining the jurisdictional amount in controversy. See Def.’s Resp. at 5.

In deciding whether diversity jurisdiction exists over a removed action, a court examines the pleading at the time of removal in order to determine the amount in controversy. See St. Paul Mercury Indem. Co. v. Red Cab Co., 303 U.S. 283, 291 (1938) (“[T]he status of the case as disclosed by the plaintiff’s complaint is controlling in the case of a removal.”); see also Anderson-Thompson, Inc. v. Logan Grain Co., 238 F.2d 598, 601-02 (10th Cir. 1956). Courts have long held that post-removal events, such as amending a complaint to reduce the damages demand to below the jurisdictional minimum, do not destroy diversity jurisdiction. See St. Paul Mercury, 303 U.S. at 292 (“And though . . . the plaintiff after removal, . . . by amendment of his pleadings, reduces the claim below the requisite amount, this does not deprive the district court of jurisdiction.”); see, e.g., Redd v. Allstate Assurance Co., No. CIV-17-246-W, 2017 WL 11139926, at *2 (W.D. Okla. Aug. 23, 2017) (“Redd’s decision to amend his state court pleading[,] . . . reduc[ing] the amount recoverable below the jurisdictional amount, will not justify remand.” (internal quotation marks omitted)).

Accordingly, because Plaintiff’s initial pleading established an amount in controversy of at least $75,000, removal based on diversity jurisdiction was proper. And, because Plaintiff’s post-removal amendment reducing the amount in controversy to less than $75,000 does not destroy that jurisdiction, Plaintiff’s request that the matter be remanded to the state court shall be denied.

Comment: If you wish to defeat efforts to remove a case, pay careful attention to the claims set forth in the initial complaint. What counts for removal is the text of the pleading on the date the case is removed.

Sims v. State Farm, Case No. CIV-23-1022-G., Western District of Oklahoma. Unpublished Opinion.

No Diversity Jurisdiction Over $0 Arbitration Award

No Diversity Jurisdiction Over $0 Arbitration Award

Tesla Motors v. Balan, 134 F.4th 558 (4th Cir. 2025) is an appeal from an effort by Tesla to confirm an arbitration award against Balan. Balan made a defamation claim and received $0. Tesla sued to confirm and the District Court confirmed the award. The Fourth Circuit reversed on the ground that there was no diversity jurisdiction. The court explained:

Because a “look through” approach is prohibited under Badgerow [v. Walters, 596 U.S. 1, 142 S.Ct. 1310 (2022)] the facts establishing a jurisdictional basis must be present on the face of the application or petition to confirm an arbitration award. See id. at 16-17, 142 S.Ct. 1310. Put differently, facts establishing that the amount in controversy exceeds $75,000 must be present on the face of a Section 9 petition to confirm an arbitration award before a district court can assert diversity jurisdiction over the action. Id. at 9, 142 S.Ct. 1310; see also Sky-Med, Inc. v. Fed. Aviation Admin., 965 F.3d 960, 965 (9th Cir. 2020).

That requirement is not satisfied in this case. Appellees went to the district court to confirm a zero-dollar award dismissing Balan’s libel claims. On its face, a petition to confirm a zero-dollar award cannot support the amount in controversy requirement. Consequently, because jurisdictional facts establishing the amount in controversy requirement are not found on the face of the petition, and a court cannot “look through” the petition to the underlying substantive controversy under Section 9, we hold that the district court did not have subject matter jurisdiction.

Comment: under the decision Tesla can confirm the award in a state court with jurisdiction over Balan. Diversity jurisdiction requirements are technical and lawyers are often surprised by just how technical they are.

Ed Clinton, Jr.

http://www.clintonlaw.net

Another One Bites the Dust

Another One Bites the Dust

Once again, a lawyer has used generative AI to draft a brief, with embarrassing consequences. In Mid-Central Operating Engineers Health and Welfare Fund v. Hoosiervac, LLC SD Indiana 2025, the 2:24-cv-00326, a lawyer got into trouble for using Generative AI to draft a brief. The court that Rule 11 was violated and that there were violations of the Indiana Rules of Professional Conduct.

On October 29, 2024, attorney Ramirez filed a brief in support of Defendant’s Motion to Reconsider the Court’s Denial of Motion to Transfer. [Dkt. 65.] In that brief, Mr. Ramirez cited to In re Cook County Treasurer, 773 F.3d 834 (7th Cir. 2014)—a case the Undersigned was unable to locate. In response to the Undersigned’s Order to file a Notice with the correct citation, [Dkt. 82], Mr. Ramirez filed a Notice in which he stated that he was unable to locate the case, “acknowledge[d] that the referenced citation was in error,” “withdr[ew] the previously cited authority[,] and apologize[d] to the Court and opposing counsel for the confusion.” [Dkt. 86 at 1.]

On December 23, 2024, the Undersigned explained that “filing a brief with a non-existent citation falls far short of an attorney’s duty to the Court, his client, and opposing counsel.” [Dkt. 87.] Accordingly, pursuant to Federal Rule of Civil Procedure 11(c)(3), the Undersigned ordered Mr. Ramirez to appear in-person and show cause why he should not be sanctioned for violating Federal Rule of Civil Procedure 11(b). Id. In that Order, the Undersigned noted that a non-exhaustive review of Mr. Ramirez’s other filings in this case revealed citations in two other briefs that the Undersigned was unable to locate: Knoedler Manufactuers, Inc. v. Cox, 545 F.2d 1033, 1035 (7th Cir. 1976), cited in [Dkt. 39]; and Brown v. Local 58, IBEW, 628 F.2d 441 (6th Cir. 1980), cited in [Dkt. 52].

On January 3, 2025, the parties in this matter appeared by counsel for a hearing on the Order to Show Cause. [Dkt. 88.] Mr. Ramirez admitted that he had relied on programs utilizing generative artificial intelligence (“AI”) to draft the briefs. Mr. Ramirez explained that he had used AI before to assist with legal matters, such as drafting agreements, and did not know that AI was capable of generating fictitious cases and citations. These “hallucination cites,” Mr. Ramirez asserted, included text excerpts which appeared to be credible. As such, Mr. Ramirez did not conduct any further research, nor did he make any attempt to verify the existence of the generated citations. Mr. Ramirez reported that he has since taken continuing legal education courses on the topic of AI use and continues to use AI products which he has been assured will not produce “hallucination cites.” Mr. Ramirez agreed during the hearing that he did not fully comply with Rule 11, but he emphasized that at no point did he act in bad faith or proceed with malice.

The court recommended a sanction of $15,000, one for each of the briefs with fake citations.

The court also found three violations of the Indiana Rules of Professional Conduct.

At least three of the Indiana Rules of Professional Conduct are implicated here, and the Undersigned will address them each in turn.

A. Rule 1.1. Competence

“A lawyer shall provide competent representation to a client. Competent representation requires the legal knowledge, skill, thoroughness and preparation reasonably necessary for the representation.” Ind. R. of Prof. Conduct 1.1. It is clear that Mr. Ramirez failed to provide competent representation when he submitted false legal bases to the Court. These actions flouted the requisite legal knowledge, skill, preparation, and especially thoroughness reasonably necessary for Mr. Ramirez’s representation of his client.

B. Rule 3.1. Meritorious Claims and Contentions

“A lawyer shall not bring or defend a proceeding, or assert or controvert an issue therein, unless there is a basis in law and fact for doing so that is not frivolous, which includes a good faith argument for an extension, modification or reversal of existing law.” Ind. R. of Prof. Conduct 3.1. This Rule mirrors the requirements of Rule 11(b)(2), and likewise proscribes presenting unfounded legal bases. There is no merit in relying on non-existent cases, so Mr. Ramirez’s conduct clearly disregards this Rule.

C. Rule 3.3. Candor Toward the Tribunal

“A lawyer shall not knowingly: . . . make a false statement of fact or law to a tribunal or fail to correct a false statement of material fact or law previously made to the tribunal by the lawyer.” Ind. R. of Prof. Conduct 3.3(a)(1). It is undisputed that Mr. Ramirez made false statements of law to the Court when in three separate submissions he relied on at least three cases which do not exist. Moreover, Mr. Ramirez did so “knowingly”—he knowingly failed to fulfill his duty of verifying that the law he presented was “good law.” As such, the Undersigned believes that discipline for violation of Indiana Rule of Professional Conduct 3.3 is appropriate as well.

Conclusions: (a) Do your own work; (b) don’t use generative AI to draft briefs because it can hallucinate; (c) even if you use AI, make sure to check the end product against real cases.

Use of Generative AI Leads To An Order To Show Cause in Federal Court

Use of Generative AI Leads To An Order To Show Cause in Federal Court

This is the case of Coomer v. Lindell, No. 22-cv-01129, pending in federal court in Colorado. Mike Lindell’s lawyers used Generative AI to respond to a brief and filed it without cite-checking the brief. This District Court entered an order to Show Cause why they should not be sanctioned. Lawyers don’t understand that Generative AI will hallucinate cases if it cannot find the citations you want it to find. This is problem with Generative AI. It is also why AI is useful. This is a path towards creativity in AI. We can’t say it enough, “Don’t Do This!”

www.clintonlaw.net

Rule 11 Does Not Apply To Pre-Removal Pleadings

Rule 11 Does Not Apply To Pre-Removal Pleadings

Cases are often filed in state court and removed to federal court by a petition for removal. Can the defendant seek sanctions under Rule 11 for the pre-removal pleadings? Most Courts say “No.”

Rule 11 provides that “[e]very pleading, written motion, and other paper must be signed by at least one attorney of record[.]” Fed. R. Civ. P. 11(a). And by presenting such a document to the Court, “whether by signing, filing, submitting, or later advocating it,” an attorney certifies that to the best of his or her knowledge it is not presented for an improper purpose, its legal contentions are warranted, and its factual contentions or denials of factual contentions have support. Fed. R. Civ. P. 11(b). If a document violates one of these certifications, then the Court may sanction the attorney, law firm, or party. Fed. R. Civ. P. 11(c).

But “Rule 11 is an extraordinary remedy, one to be exercised with extreme caution.” Mason Classical Acad., Inc v. Arnn, No. 2:22-CV-513-JLB-NPM, 2023 WL 6196855, at *5 (M.D. Fla. Sept. 22, 2023) (citation omitted). This extraordinary remedy is inappropriate at this juncture for two reasons.

First, when a case is removed from state court (like here), “Rule 11 does not apply to pleadings filed before removal.” Worldwide Primates, Inc. v. McGreal, 26 F.3d 1089, 1091 (11th Cir. 1994). JM Family Defendants’ sanctions motion relies exclusively on the amended complaint filed in state court. (See, e.g., Doc. 48 at 13, 16 (reciting paragraphs of the amended complaint)). And, while post-removal advocacy can support Rule 11 sanctions, the Defendants’ motion does not rely on any of Plaintiff’s post-removal filings. This is problematic, as Rule 11 mandates that the sanction motion “describe the specific conduct that allegedly violates Rule 11(b).” Fed. R. Civ. P. 11(c)(2) (emphasis added).

The case is Lynch v. Filice, District. Court MD Florida 2025 2-24-cv-340-SPC-NPM. It is unpublished.

Edward X. Clinton, Jr.

No Case Ending Sanctions Where Plaintiff Sought to Comply

No Case Ending Sanctions Where Plaintiff Sought to Comply

In Greer v. Moon, District. Court, D. Utah 2025, the court denied a defense request for Rule 37 sanctions because the pro se litigant amended his disclosures in response to the court’s prior orders. The plaintiff’s Rule 26 disclosures were deficient.

The court explains: As to the first Woodworker’s factor, although Mr. Greer’s failure to properly disclose witnesses or evidence to Defendants is not harmless and has resulted in needless litigation and expenses, Defendants have not suffered the prejudice or surprise that Rule 37(c) exclusion contemplates, such as the introduction of new witnesses after the close of discovery, after expert reports and depositions are completed, and after dispositive motions have been filed.[11] Instead, the court is able to rectify Defendants’ needless expenditures for needless litigation by ordering Mr. Greer to pay those expenses. If Mr. Greer later attempts to introduce information or witnesses he failed to produce in initial disclosures, the court may address Rule 37(c) exclusion at that time. Presently, incurring expenses does not hinder Defendants from presenting their defense and, as discussed below, is curable by methods other than exclusion.

However, the court denied “case-ending” sanctions.

The court also declines to impose the “case-ending” sanctions Defendants request for Mr. Greer’s failure under Rule 26. The court warned Mr. Greer that if he failed to comply with the court’s order compelling production of the Rule 26(a) information, he may be subject to potential case-ending sanctions under Rule 37(b)(2).[16]Dismissing this case when costs and fees would rectify prejudice to Defendants would violate the clear preference to adjudicate cases on their merits.[17] That is especially true where, as here, Mr. Greer is proceeding pro se. Albeit imperfect, Mr. Greer has attempted to comply with the court’s order to provide initial disclosures to Defendants, which renders Rule 37(b)(2) sanctions unnecessary at this juncture.

Comment: pro se litigants are governed by the same rules everyone else is, but they have a more difficult time complying with those rules.