Nine of 27 citations were wrong and two cases did not exist: a special master ordered Ellis George and K&L Gates to pay $31,100

This is an honest-negative AI-governance case, not a deployment win. Every figure originates with an independent court officer, not a self-reporting vendor or buyer.

The problem

Generative AI marketed for legal research fabricates authority, and the failure does not stay inside the firm that used it. In Lacey v. State Farm General Insurance Co. (C.D. Cal., No. 2:24-cv-05205-FMO-MAA), a court-appointed special master had been asked to resolve a discovery dispute over the insurer’s privilege assertions (source). Plaintiff’s supplemental brief on that issue “contained numerous false, inaccurate, and misleading legal citations and quotations” (source). On the special master’s after-the-fact review, “approximately nine of the 27 legal citations in the ten-page brief were incorrect in some way. At least two of the authorities cited do not exist at all. Additionally, several quotations attributed to the cited judicial opinions were phony and did not accurately represent those materials” (source).

The order is explicit about why this mattered rather than merely embarrassed (source):

“Directly put, Plaintiff’s use of AI affirmatively misled me. I read their brief, was persuaded (or at least intrigued) by the authorities that they cited, and looked up the decisions to learn more about them – only to find that they didn’t exist. That’s scary. It almost led to the scarier outcome (from my perspective) of including those bogus materials in a judicial order.”

What was built

No system was built. What failed was a hand-off (source). The order records it directly (source):

“The lawyers admit that Mr. Copeland, an attorney at Ellis George, used various AI tools to generate an “outline” for the supplemental brief. That document contained the problematic legal research.”

He sent the outline to lawyers at co-counsel K&L Gates, who “incorporated the material into the brief” (source). The order records the control gap in one sentence: “No attorney or staff member at either firm apparently cite-checked or otherwise reviewed that research before filing the brief with the Special Master” (source). The receiving firm “didn’t know that Mr. Copeland used AI to prepare the outline; nor did they ask him” (source).

The order also records a second failure after the problem surfaced. When the special master emailed about two citations he could not confirm, “K&L Gates re-submitted the brief without the two incorrect citations – but with the remaining AI-generated problems in the body of the text” (source). An associate attorney then wrote back; the order’s sentence describing that e-mail straddles the page-3/4 footnote gutter in the primary, and the part that runs contiguously on page 4 reads (source):

“e-mail thanking me for catching the two errors that were “inadvertently included” in the brief, and confirming that the citations in the Revised Brief had been “addressed and updated.””

The order does not name any AI product; it attributes the fabrications to “AI tools” generically (source). Third-party reporting names specific tools, but that attribution is deliberately held out of this record until it can be tied to a quoted line in the primary (source).

The outcome

The special master found the lawyers “collectively acted in a manner that was tantamount to bad faith” and that “[t]he initial, undisclosed use of AI products to generate the first draft of the brief was flat-out wrong” (source).

The number that gets quoted: $31,100, jointly and severally, against Ellis George and K&L Gates. “Plaintiff’s law firms are ordered (jointly and severally) to pay compensation to the defense in the aggregate amount of $31,100” (source), a figure reported the same way by two mutually independent newsrooms (source, source).

The scale of the fabrication: 9 of 27 citations wrong, at least 2 non-existent. The judge found roughly a third of the citations in the ten-page brief incorrect and at least two of the cited authorities fictional (source, source).

The briefs struck and the discovery relief denied. The special master “imposed the monetary penalty, struck the deficient brief, and barred further discovery on the issue that it addressed” (source); in the order’s own words, “I have struck, and decline to consider, any of the supplemental briefs that Plaintiff submitted on the privilege issue” and “I decline to award any of the discovery relief […] that Plaintiff sought” (source).

No sanction on the individual attorneys. “In a further exercise of discretion, I decline to order any sanction or penalty against any of the individual lawyers involved here,” the special master wrote, reasoning that “[t]his was a collective debacle, and is properly resolved without further jeopardy” (source). The ABA Journal confirms it independently: “Wilner did not sanction the individual attorneys involved in the mishap, however” (source). The money falls on the firms, not the client: “Mrs. Lacey is clearly not at fault for the AI debacle […] She will not, however, be financially responsible for the monetary awards described in this order” (source).

The $31,100 breaks into approximately $26,100 reimbursing special-master fees the defendant had advanced, plus $5,000 toward the defense’s own fees (source). This split is a weaker claim than the total: it appears in the primary and in Reason’s Volokh Conspiracy, which prints it only by block-quoting the order, and neither the ABA Journal nor the LawSites report carries either figure, so it rests on the order and its reprint rather than on independent corroboration.

How this was verified

This case carries a green badge under TIN’s court-origin standard: the origin is a signed order of a court-appointed special master, an independent court officer, not a self-reporting vendor or buyer. The order (Docket #119, C.D. Cal., signed 5 May 2025, filed 6 May 2025) was fetched from the CourtListener RECAP mirror and byte-tied to a Wayback capture with an identical sha1. On 2026-08-16 the headline figures were re-checked against three live sources: Reason’s Volokh Conspiracy, which reproduces the order verbatim, plus two mutually independent newsrooms, the ABA Journal (Debra Cassens Weiss) and LawSites (Bob Ambrogi). The $31,100 total, the 9-of-27 citation count, the struck briefs, and the no-individual-sanction ruling all re-verified. The one figure carried at a weaker standard is the $26,100/$5,000 split, which no independent secondary reports and which is presented above with that caveat. The docket-currency check is closed: the case settled (entry #145, 23 July 2025) and was dismissed (entry #146, 24 July 2025), with no notice of appeal, so the order stands as issued. No AI product is named in the record, and no claim here asserts one.

Related case files

The same reputable-firm-plus-unverified-AI failure recurs in Johnson v. Dunn, where Butler Snow lawyers filed ChatGPT-hallucinated citations and were disqualified, showing the disciplinary end of the same conduct that here drew only a fee award. Wadsworth v. Walmart is the closest structural sibling: a large plaintiff-side firm sanctioned under Rule 11 for the identical control gap, an AI-generated draft filed without cite-checking. Gauthier v. Goodyear is the earlier, smaller marker that set the duty Lacey enforces: a lawyer remains responsible for verifying AI output before filing.

---

Sources

Checked live on 2026-08-16. Tier 1 = the primary court order; Tier 2 = independent reputable press naming the parties. The order does not name any AI tool, so none is asserted here.

1. Special Master, “Order of Special Master Imposing Non-Monetary Sanctions and Awarding Costs,” Lacey v. State Farm General Insurance Co., No. 2:24-cv-05205-FMO-MAA, Doc. #119 (C.D. Cal., signed 2025-05-05, filed 2025-05-06) (Tier 1, primary court order). https://websitedc.s3.amazonaws.com/documents/Lacey_v._State_Farm_General_Insurances_Co._D._Cal._May_6_2025.pdf
2. Reason / The Volokh Conspiracy, “AI Hallucination in Filings Involving 14th-Largest U.S. Law Firm Lead to $31K in Sanctions,” 2025-05-13 (Tier 2, reproduces the order verbatim). https://reason.com/volokh/2025/05/13/ai-hallucination-in-filings-involving-14th-largest-u-s-law-firm-lead-to-31k-in-sanctions/
3. ABA Journal (Debra Cassens Weiss), “Sanctions imposed for ‘collective debacle’ involving AI hallucinations and 2 firms, including K&L Gates,” 2025-05-14 (Tier 2, independent narration). https://www.abajournal.com/web/article/judge-imposes-sanctions-for-collective-debacle-involving-ai-hallucinations-and-2-law-firms-including-k
4. LawSites (Bob Ambrogi), “AI Hallucinations Strike Again: Two More Cases Where Lawyers Face Judicial Wrath for Fake Citations,” 2025-05 (Tier 2, independent narration). https://www.lawnext.com/2025/05/ai-hallucinations-strike-again-two-more-cases-where-lawyers-face-judicial-wrath-for-fake-citations.html