On May 9, 2026, U.S. District Judge Nina Y. Wang issued a $5,000 Rule 11 sanctions order in Coomer v. Lindell, the latest blow in a fast-moving wave of attorney discipline tied to AI-generated fake citations. The penalty fell jointly on attorney Christopher Kachouroff, his firm, co-counsel Jennifer DeMaster, and FrankSpeech for presenting Capital Solutions, LLC v. Konica Minolta Business Solutions USA, Inc., a federal district court decision, as if it were a Tenth Circuit ruling on punitive damages and the Seventh Amendment Reexamination Clause. Kachouroff disclaimed any generative-AI use in producing the brief and pointed to Westlaw, but the court expressed open skepticism given the recurrence of hallucination-style errors.
The May order was Wang's second sanctions ruling against the same lawyers in the same case. In July 2025, she had already ordered Kachouroff and DeMaster to pay $3,000 each after identifying roughly 30 defective citations in earlier filings. Local coverage in Colorado Politics documented the run-up to the May ruling, noting prior warnings from the bench and the underlying $2.3 million verdict in Eric Coomer's favor.
Why criminal defenders carry a heavier risk than civil counterparts
Civil litigants face money damages. Criminal defendants face liberty. That difference reshapes the consequences of an AI-fabricated citation in three concrete ways.
First, a phantom case in a suppression motion can forfeit the underlying argument entirely. If a court rejects a Fourth Amendment challenge because the supporting authority does not exist, the issue is gone, and so is the suppression remedy. Second, fake authority in a sentencing memorandum or post-conviction brief misleads the tribunal in proceedings where the stakes are years of incarceration. Third, the same conduct seeds an ineffective-assistance claim under Strickland v. Washington: a defendant whose lawyer filed hallucinated case law has a credible argument that counsel's performance fell below an objective standard of reasonableness.
The candor duty is identical across civil and criminal practice. ABA Model Rule 3.3 forbids knowingly making false statements of law to a tribunal, and Rule 1.1 demands competent representation. ABA Formal Opinion 512, issued July 29, 2024, applies both rules directly to generative AI: lawyers must understand the limits of any GAI tool they use and must perform task-specific independent verification of every output before filing.
The 2026 sanctions wave
Coomer v. Lindell is one entry in a docket that expanded sharply through 2026. Researcher Damien Charlotin's hallucination database had logged more than 1,200 instances worldwide of AI-fabricated citations submitted to courts as of early 2026, growing by roughly 500 entries in three months and adding five to six new documented filings per day. U.S. courts imposed more than $145,000 in AI-related sanctions in the first quarter of 2026 alone, according to practitioner trackers compiling sanction taxonomies.
Recent orders that criminal defenders should know:
-
Rivera v. Triad Properties Corp. (N.D. Ala., April 2026): Judge Anna Manasco imposed $47,056.90 in fees against Joshua B. Watkins and Burrill Watkins LLC under Rule 11 and the court's inherent authority, citing AI-fabricated citations and what she called "feigned contrition." A bar-counsel referral accompanied the financial sanction, illustrating the current top end of U.S. sanctions for this conduct.
-
Whiting v. City of Athens (6th Cir., March 2026): A federal appeals court sanctioned Tennessee attorneys Van R. Irion and Russ Egli for filing briefs containing more than two dozen fabricated or misrepresented citations. It is one of the highest-court AI-sanctions orders to date.
-
Kenosha County, Wisconsin (February 2026): Circuit Judge David Hughes struck a brief filed by District Attorney Xavier Solis that contained undisclosed AI-generated fake citations. Charges against two defendants, totaling 74 counts including 38 felonies, were dismissed without prejudice. The episode is a vivid example of how a hallucination can change the posture of a criminal case in a single ruling.
-
Mata v. Avianca, Inc. (S.D.N.Y. June 22, 2023): The foundational precedent. Judge P. Kevin Castel imposed $5,000 in Rule 11 sanctions on Steven Schwartz and Peter LoDuca after they filed ChatGPT-fabricated cases, finding "subjective bad faith." Every later AI-hallucination order traces back to this opinion.
Discipline beyond fines
Money sanctions are the floor, not the ceiling. Colorado regulators suspended attorney Zachariah C. Crabill for two years, with 90 days served and the balance probationary, after he filed ChatGPT-generated fake cases. The Crabill order is the leading bar-discipline precedent for AI hallucinations and is now routinely cited by regulators arguing for proportionality at the suspension end of the spectrum. The Rivera court's referral to bar counsel signals that fee shifts and disciplinary investigations are increasingly bundled together rather than imposed as alternatives.
Bar-rule reform is moving fast
On January 8, 2026, the Colorado Supreme Court became the first U.S. jurisdiction to adopt AI-specific amendments to its Rules of Professional Conduct, formalizing competence, supervision, and verification duties around generative AI. Other state bars have issued parallel guidance, including the New York State Bar Association's guardrails for lawyers. A growing number of federal trial judges now require AI-use disclosures or certifications in standing orders, and the trend points toward formal local-rule amendments in additional districts.
A verification workflow for criminal defense practice
Every published court order on AI hallucinations rests on the same finding: counsel did not perform a reasonable inquiry. Rule 11(b)(2) and the courts' inherent authority converge on a workable practice standard.
-
Never copy a citation, quotation, or holding directly from a chatbot output. Treat AI text as an unverified draft, nothing more.
-
Pull every cited case in Westlaw or Lexis. Read at least the headnote and the cited pinpoint page.
-
Confirm the deciding court and the reporter. The Capital Solutions error in Coomer turned on a federal district court opinion misrepresented as Tenth Circuit authority.
-
Verify quotations against the source opinion, not against an AI summary of the opinion.
-
Maintain a citation-verification log in the case file. The log is the work product that proves a reasonable inquiry occurred under Rule 11.
-
Require a second-pair review for any motion implicating constitutional rights, suppression issues, or sentencing exposure.
-
Comply with any standing-order disclosure requirements in the trial or appellate court.
-
Supervise paralegals, contract attorneys, and outside drafters consistent with Model Rules 5.1, 5.2, and 5.3.
If a hallucination slips through
The mitigation playbook is narrow but real. Withdraw and correct the filing promptly. Disclose the error candidly to the court rather than waiting for opposing counsel to surface it. Preserve the work product showing the verification effort that was attempted. Self-report to bar counsel where the rules require it. Courts have weighed these steps when calibrating sanctions, and the contrast between the Mata defendants' delayed admissions and the Rivera court's "feigned contrition" finding is instructive. Sanction severity tracks well-defined factors: number of fabricated authorities, presence of false quotations, continuation after notice, candor in admitting error, remorse, and prejudice to the proceeding. All of those factors are amplified in criminal practice, where liberty interests are at stake.
From rebuke to disbarment
The trajectory in the published orders runs in one direction. A first incident draws a $5,000 Rule 11 sanction in line with Mata. A repeat or aggravated incident produces a $47,000 fee shift and a bar-counsel referral, as in Rivera. Patterned conduct in the same case yields a second sanctions order, as Judge Wang issued in Coomer v. Lindell. Discipline can escalate to a multi-year suspension on the Crabill model, and disbarment for repeat or knowing offenders is a foreseeable next step. Independent verification of every AI-derived citation is no longer a best practice. It is a baseline competence obligation under the rules now in force.
Related reading
Sources
-
Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023) (Justia)
-
ABA Formal Opinion 512: Generative AI Tools (ABA, July 29, 2024)
-
The 2026 AI-Related Amendments to the Colorado Rules of Professional Conduct (Colorado Lawyer)
-
AI Hallucinations in Court Filings: A 2025 Review of Sanctions Across the Courts (Sterne Kessler)
-
Avoiding Sanctions in the Gen AI Era: Practical Guardrails for Lawyers (NYSBA)
-
Colorado Atty Suspended For Using 'Sham' ChatGPT Case Law (Federal Defender Services Office)
Note: This article contains AI-assisted content and has been reviewed by our editorial team.
IMPORTANT NOTICE: GavelDaily.com is an advertising and information service, not a law firm. Submitting this form does not create an attorney-client relationship. The information you provide will be shared with licensed criminal defense attorneys in your area who may contact you about your inquiry. All submissions are treated as confidential by our service, but attorney-client privilege does not attach until you formally retain an attorney. Any attorney you connect with through this service is independently responsible for their legal advice and representation. Free consultations are subject to individual attorney availability. If you are facing an immediate legal emergency, contact your local public defender's office or call 911.