Reported / Citable
Background
Plaintiff Matt McCormick brought a Telephone Consumer Protection Act suit against Texakoma Financial, Inc., alleging the company called his phone number three times despite its registration on the National Do Not Call Registry. When Texakoma moved for summary judgment in December 2025, plaintiff’s counsel Amy L.B. Ginsburg filed a response brief in January 2026. Texakoma’s reply flagged that the brief cited a case that did not appear to exist, along with fictitious quotations and propositions unsupported by the real cases cited. Ginsburg did not file a sur-reply or take any steps to correct the record.
After independently verifying Texakoma’s accusations, the court issued an order to show cause in May 2026, directing Ginsburg to appear in person and bring highlighted copies of the cases she had cited. Less than one hour before the hearing, Ginsburg filed a written response attributing the defective brief to a law clerk or intern who may have used AI, and to a paralegal who allegedly filed an uncorrected earlier draft. She claimed to have personally identified the fictitious Noviello v. Bayview Asset Management citation during her review and to have corrected it, but could not produce any revised draft, metadata, or corroborating evidence.
At the show cause hearing, Ginsburg acknowledged the citations were problematic, conceded the fictitious case and fabricated quotes were sanctionable under Rule 11, and expressed remorse — while maintaining she would never knowingly submit false citations. Defense counsel reported incurring approximately $9,000 preparing the reply brief and $2,000 attending the hearing.
The Court’s Holding
Judge Mazzant found that Ginsburg’s brief contained three categories of serious Rule 11 violations: (1) a wholly fictitious case — Noviello v. Bayview Asset Management, No. 3:18-cv-02057 — whose case number and Westlaw reporter number each lead to entirely different, unrelated matters; (2) two fabricated quotations attributed to the real case Texas v. American Blastfax, Inc., 164 F. Supp. 2d 892 (W.D. Tex. 2001); and (3) six propositions attributed to real cases that those cases do not actually support. The court rejected Ginsburg’s explanation as not credible, finding her account a “contrived attempt to shift the blame to others” that raised more questions than it answered and was undermined by her equivocal, minimizing language and inability to produce any evidence of a corrected draft.
The court concluded that a reprimand alone, or financial sanctions alone, would be insufficient deterrents given the national surge in AI-hallucination incidents — noting that a tracked database of attorney hallucination cases grew from 239 (as of February 2026) to 434 by the date of this order. Applying Rule 11(c)(4)’s requirement that sanctions be “limited to what suffices to deter repetition of the conduct or comparable conduct by others similarly situated,” the court imposed a combination of monetary and non-monetary sanctions.
Ginsburg was publicly reprimanded, and the court ordered the Clerk to serve the order on every district and magistrate judge in the Eastern District of Texas. Ginsburg was further ordered to conduct a reasonable review of every filing bearing her signature from January 1, 2026 through the date of the order in any federal or state court, verify all cited authorities, correct any misrepresentations within three days of discovery, and file a sworn certification of compliance within ninety days. The court also left open the possibility of a fee-shifting motion by defense counsel, to be filed within fourteen days, with instructions to meet and confer on whether fees were reasonably necessary.
Key Takeaways
- A fictitious case citation, fabricated quotations attributed to real cases, and propositions unsupported by cited authority each independently violate Rule 11(b)(2), regardless of whether counsel personally generated the errors using AI or delegated drafting to a subordinate.
- The “wrong version filed” defense — blaming a paralegal for uploading an uncorrected draft — carries no weight without supporting evidence such as metadata, version history, or a recoverable revised file; courts are increasingly aware that this explanation has been deployed repeatedly to avoid AI-sanctions.
- Eastern District of Texas Local Rule AT-3(m) explicitly requires attorneys who use generative AI to review and verify all content, and the court treated Ginsburg’s familiarity with AI-related CLE training as a reason to be more — not less — skeptical of her claimed ignorance.
- Financial sanctions alone are no longer viewed as an adequate deterrent for AI-hallucination misconduct; courts are moving toward non-monetary consequences, including public reprimands, mandatory audits of prior filings, and sworn certifications.
- The client was not sanctioned and the court allowed a corrected summary judgment response to be filed, insulating the plaintiff from prejudice caused by counsel’s misconduct.
Why It Matters
This decision is one of a rapidly growing line of cases in which courts are escalating sanctions beyond fines in response to AI-generated citation fraud. Judge Mazzant’s order is notable for its breadth: the mandatory backward-looking audit of all of Ginsburg’s 2026 federal and state court filings is among the more expansive corrective remedies imposed to date, and the court’s explicit reliance on a publicly tracked database showing 434 hallucination incidents nationwide signals that judges are now treating this as a systemic crisis rather than an isolated lapse.
For practitioners, the opinion underscores that supervisory responsibility for subordinate-drafted work is non-delegable. An attorney who signs a brief certifies its contents under an objective standard of reasonableness, and that standard requires actual verification of every cited authority — not an assumption that a clerk or paralegal handled the problem. As AI drafting tools become more common, this case reinforces that the failure to verify is itself the violation, irrespective of who first generated the erroneous text.