The American Bar Association Standing Committee on Ethics and Professional Responsibility issued Formal Opinion 512 addressing the use of artificial intelligence:  'To ensure clients are protected, lawyers using generative artificial intelligence tools must fully consider their applicable ethical obligations, including their duties to provide competent legal representation, to protect client information, to communicate with clients, to supervise their employees and agents, to advance only meritorious claims and contentions, to ensure candor toward the tribunal, and to charge reasonable fees.'

As the Sixth Circuit recently observed: 'New technologies present significant promise for the legal field.  But all in the legal profession must be clear eyed about technology’s potential pitfalls. That mandate is especially critical in today’s rapidly evolving artificial-intelligence landscape . . . . Attorneys should not utilize technology without knowing the ways in which it can be misused or contribute to inaccuracies. That remains true even when new tools are sponsored by trusted legal technology providers.' United States v. Farris, 171 F. 4th 920, 922 (6th Cir. 2026).

In civil litigation, some district courts are imposing significant monetary penalties for the misuse of AI, often in the form of the offending litigant paying opposing counsel’s fees for hours related to addressing the hallucinations in the filing. The first widely reported AI sanctions case was in 2023, Mata v. Avianca, 678 F. Supp. 3d 443 (S.D.N.Y.). In ByoPlanet International v. Johansson, 792 F. Supp. 3d 1341 (S.D. Fla. 2025), the court ordered an attorney (who admitted he uncritically relied on GenAI to do his work) to pay over $85,000 in attorneys’ fees. See also, e.g., Mattox v. Prod. Innovations Rsch., LLC, 807 F. Supp. 3d 1341 (E.D. Okla. 2025) (imposing a total sanction of $23,495.90); Lacey v. State Farm Gen. Ins. Co., 2025 WL 1363069, at *5 (C.D. Cal. May 5, 2025) (ordering Plaintiff’s law firms, jointly and severally, to pay defendant’s attorneys’ fees of $31,000). 

If an expert witness improperly relies on AI to develop their findings, the expert’s credibility is undermined and their testimony may be excluded. In Kohls v. Ellison, the district court excluded testimony from an AI expert witness who included fake, AI-generated sources in his declaration. 2025 WL 66514, at *4 (D. Minn. Jan. 10, 2025).

Revocation of pro hac vice status is another sanction used when lawyers misuse GenAI. Because 'society’s interest in avoiding the unethical behavior of citing fictitious cases is significant,' revocation of pro hac vice status may be appropriate for attorney misconduct based on the failure to verify work produced by GenAI. See Wadsworth v. Walmart, 348 F.R.D. 489, 497-98 (D. Wyo. 2025)

Courts have referred attorneys to the appropriate disciplinary body for the misuse of GenAI. See, e.g., Doc App., Inc. v. Leafwell, Inc., 2025 WL 3288251, at *6 (M.D. Fla. Nov. 26, 2025) (referring counsel to the Florida bar); In re Kheir, 674 B.R. 631, 633 (Bankr. S.D. Tex. 2025) ('[Attorney] King will also be referred to Chief United States District Judge Randy Crane and the State Bar of Texas Chief Disciplinary Counsel for possible disciplinary action.'). Judges may also require attorneys to self-report their misuse of GenAI to the appropriate attorney disciplinary body. See Safe Choice, LLC v. City of Cleveland, 2025 WL 3029553, at *6 (N.D. Ohio Oct. 30, 2025). In a particularly egregious misuse case, three attorneys were referred to the Alabama State Bar, publicly reprimanded, disqualified from the case, and ordered to provide a copy of the sanctions order to all parties and judges involved in all cases in which they were counsel of record. Johnson v. Dunn, 792 F. Supp. 3d 1241 (N.D. Ala. 2025).

Mandated training about artificial intelligence is a common sanction, but it is typically imposed in conjunction with other sanctions. See, e.g., In re Kheir, 674 B.R. 631, 633 (Bankr. S.D. Tex. 2025) (ordering attorney King to pay opposing counsels’ attorneys’ fees, register and obtain six hours of CLE on the use of GenAI, and provide a copy of the order to his client along with a certificate of compliance once he has done so); Lafontant v. Coolidge-CLK St. Germane, LLC, 2025 WL 2817583, at *3 (E.D. La. Oct. 3, 2025) (referring attorney to the Eastern District of Louisiana’s Disciplinary Committee, ordering them to attend CLE on GenAI, and fining them $1,000 for filing a brief in opposition to a motion for summary judgment containing two cases); Versant Funding LLC v. Teras Breakbulk Ocean Navigation Ents., LLC, 2025 WL 1440351, at *7 (S.D. Fla. May 20, 2025) (ordering counsel to pay plaintiff’s attorneys’ fees and mandating completion of 'an approved CLE on Artificial Intelligence which has an ethics component addressing the ethical and legal obligations of counsel' within thirty days of the order).

As the number of case filings with hallucinated citations has increased, many judges have increased the range and severity of sanctions, particularly when lawyers make repeated AI-generated errors. Two attorneys were held in contempt of court after their 'continued malfeasance le[ft] the [c]ourt almost at a loss for words' based on their repeated, flagrant misuse of AI in numerous filings. See Kettering Adventist Healthcare v. Collier, 2026 WL 523117, at *1 (S.D. Ohio Feb. 25, 2026). 

In some cases, district courts have struck pleadings or dismissed cases. Two attorneys were referred for discipline and their client’s complaint was dismissed as a sanction for including citations to 'quotes that d[id] not appear in binding Eleventh Circuit opinions' and a decision from that district '[c]ourt that d[id] not exist.' Dubinin v. Papazian, 2025 WL 3248187, at *1-3 (S.D. Fla. Nov. 21, 2025). Another attorney, after being ordered to show cause for including hallucinated case law in a brief, filed a response with more hallucinated citations. This resulted in the court striking his filings and entering a default judgment against his client. Flycatcher Corp. v. Affable Ave. LLC, 2026 WL 306683, at *1 (S.D.N.Y. Feb. 5, 2026). 

Hallucinated cases have also appeared in pro se filings. However, most courts issue a warning to a pro se litigant before imposing sanctions. In some cases, pro se litigants have explained that because they lack access to traditional legal databases, they relied on GenAI to fill that gap. See Kruglyak v. Home Depot U.S.A., Inc., 774 F. Supp. 3d 767 (W.D. Va. 2025).

However, courts have imposed significant sanctions on pro se litigants who continue to submit filings with hallucinations after several warnings, especially if they do so in response to a show cause order based on AI misuse. See Ferris v. Amazon Servs., LLC, 778 F. Supp. 3d 879 (N.D. Miss. 2025). A pro se litigant had her case dismissed with prejudice after she used AI to both falsify evidence and draft her pleadings, including citing a fake Fourth Circuit case to support her claim that AI use was permissible. Robinson v. Oglala Sioux Tribe, 2025 WL 2609573, at *3-4 (W.D. Okla. Sep. 9, 2025). Another pro se litigant had his filing struck after the district court found he relied on a third party to draft and file his motion that included several 'fabricated arguments generated by artificial intelligence.' Andre v. Warden, FCI Danbury, 2025 WL 3281732 (D. Conn. Nov. 25, 2025). 

Courts also have sanctioned the misuse of AI in criminal cases, though the process may raise other concerns when an attorney is appointed as counsel pursuant to the Criminal Justice Act (CJA), 'a publicly funded program.' United States v. Farris, 171 F. 4th 920, 923 (6th Cir. 2026). An attorney appointed under the CJA submitted a motion riddled with hallucinated citations, including 'precedent that the Supreme Court reversed more than 40 years ago.' Sanctions included a referral to relevant disciplinary authorities, a recommendation to be removed from the CJA panel, and a $5,000 fine to the CJA fund to 'offset some of the losses.' United States v. McGee, 806 F. Supp. 3d 1264, 1267 (S.D. Ala. 2025). A federal public defender was sanctioned for including fake case law in his filings, though he never admitted using AI. See United States v. Hayes, 763 F. Supp. 3d 1054 (E.D. Cal. 2025). As of May, 2026, there are no reported cases of a federal prosecutor being sanctioned for misusing AI, but state courts are beginning to confront this issue. The Georgia Supreme Court sanctioned an assistant district attorney for AI misuse that 'sidetracked' the court 'from [its] obligation of resolving the merits' of the criminal appeal, admonishing the ADA and suspending her from practice before the court. See Payne v. State, __ S.E.2d __, 2026 WL 1215905, at *1 (Ga. 2026).

Hallucination-filled filings are appearing at the appellate level. See, e.g., Whiting v. City of Athens, 170 F. 4th 455, 466 (6th Cir. 2026) (imposing a 'substantial penalty' on attorneys for 'br[inging] the [legal] profession into disrepute' by including 27 non-existent or inaccurate case and record citations); Dec v. Mullin, 171 F. 4th 940, 947-48 (7th Cir. 2026) (admonishing attorney who denied using AI yet could not explain why her brief included 'two cases that d[id] not exist and recount[ed] an imaginary quotation'). 'AI-hallucinated case citations have increasingly become an even greater problem in [appellate] courts, and the problem shows no sign of abating.' Fletcher v. Experian Info. Sols., 168 F.4th 231 (5th Cir. 2026).