← All case files
pending deployment legal-services · US · cross

An Ohio federal judge fines two lawyers $7,500 for AI-fabricated citations — 'the most egregious' Rule 11 violation in his 46 years on the bench

In Kettering Adventist Healthcare v. Collier (No. 3:25-cv-273, S.D. Ohio), Senior Judge Walter H. Rice found attorneys Mary T. Scott and H. Leon Hewitt in contempt of court and in violation of Federal Rule of Civil Procedure 11(b), imposing a combined $7,500 in sanctions — $5,000 against Scott and $2,500 against Hewitt — payable to the Clerk of Court within thirty days. Their briefs cited at least twelve nonexistent or wholly-inapposite cases; Scott admitted she had used generative AI 'for preliminary case identification.' The court called it the most egregious Rule 11 violation it had seen in forty-six years on the federal bench, reported the conduct to the Supreme Court of Ohio Office of Disciplinary Counsel, struck the motion to dismiss, and recused.

MetricBeforeAfter
Combined $7,500 monetary sanction under Fed. R. Civ. P. 11(c) — $5,000 against Mary T. Scott, Esq. and $2,500 against H. Leon Hewitt, Esq. — payable to the Clerk of Court within thirty (30) days of entry
Both attorneys held in contempt of court and in violation of Rule 11(b); conduct reported to the Supreme Court of Ohio Office of Disciplinary Counsel
Predicate finding: at least twelve cited cases either do not exist or were so far afield they had no place in a court filing; Scott admitted using generative AI 'for preliminary case identification'
The court called it the most egregious Rule 11 violation it had seen in forty-six years on the federal bench

Verification status: PENDING — checker-graduated at confidence 1.000, awaiting HUMAN client-confirmation of docket posture. Not verified; no green badge until Maneesh confirms.

The problem

Kettering Adventist Healthcare (d/b/a Kettering Health Network) sued Sandra Collier and Mary T. Scott in the U.S. District Court for the Southern District of Ohio. Scott — a lawyer herself — appeared both pro se and as former counsel for co-defendant Collier; H. Leon Hewitt appeared as Collier’s current counsel. In their Motion to Dismiss, a Reply in support of it, and a memorandum contra the third-party defendants’ motion, the two relied on case citations that did not hold up: when the court checked them, it found the cited cases either did not exist or stood for nothing like the propositions asserted.

What was built

No system was built or deployed. This is a court-origin AI-governance honest-negative from the attorney-sanction vein. In her Jan. 15, 2026 response to the court’s show-cause order, Scott “state[d] that she used generative artificial intelligence” — in her words, “for preliminary case identification” — and conceded the resulting citations were, as the order puts it, “AI hallucinations.” The court is careful about attribution: in its Jan. 2, 2026 Decision it wrote that it “does not know whether Scott and Hewitt used AI to generate caselaw,” and that whether they did “is ultimately unimportant.” No specific product (ChatGPT, Claude, or otherwise) is named anywhere in the record; the violation rests on the duty of candor and Rule 11(b), not on a finding that any particular AI tool was used.

The outcome

On Jan. 2, 2026 (Doc. #113), Senior Judge Walter H. Rice detailed the fabrications: “Scott and Hewitt have cited at least twelve cases that either do not exist or are so far afield from the” subject matter that “they had no place in a Court filing.” Even after Kettering put them on notice, the court found, they “cited those very same cases for the very same propositions of law in their later Reply.” Reason’s Volokh Conspiracy, which reproduced that decision, recorded the same finding verbatim.

On Feb. 25, 2026 (Doc. #122), the court found both attorneys in contempt and in violation of Rule 11(b) and, pursuant to Rule 11(c), imposed monetary sanctions — $5,000 against Scott and $2,500 against Hewitt — “payable to the Clerk of Court within thirty (30) days of entry.” As the ABA Journal reported, the court “imposed a collective sanction of $7,500 against two attorneys for AI hallucinations”; as LexisNexis’s Mealey’s reported, the judge found “the lawyers in contempt of court, imposing $7,500 in sanctions and referring them for discipline.” The court set that combined figure against the “in excess of seventy-five hours” it had spent on the issue. It called the conduct “the most egregious violations of Rule 11 the undersigned has seen in his forty-six years on the federal bench” — a line Mealey’s rendered as the “most egregious, inexplicable, and repeated violations of Rule 11(b)” — reported the conduct to “the Ohio Supreme Court’s Office of Disciplinary Counsel,” struck the motion to dismiss without prejudice, and recused.

Conflation guards

  • $7,500 is the COMBINED figure. Per attorney it is $5,000 (Scott) + $2,500 (Hewitt). The court framed the combined sum as roughly $100 per hour of judicial time.
  • No AI product is named. Scott admitted generative-AI use for “preliminary case identification,” but the record identifies no specific tool; coverage attributing the citations to “ChatGPT” would not be supported.
  • “46 years” vs “55 years” reconcile. Doc. #122 says “forty-six years on the federal bench” (Rice was appointed in 1980); Doc. #113 and the Volokh headline say he has “served continuously as a trial judge since 1969,” counting his earlier state judgeship. Same judge, two spans — not a discrepancy.
  • The underlying merits are unrelated. This is a healthcare employment/whistleblower dispute; the $7,500 is the AI-citation sanction, the only monetary figure at issue here.

Path to green

Every figure is established by the court’s own record and needs no client call. The open items are docket-only and asserted by no claim: whether the attorneys paid, any reconsideration or appeal, and whether the Chief Judge (whom Rice asked to weigh a further contempt sanction) adds anything — none of which disturbs the $7,500 imposed here. A green badge — if TIN carries court-adjudicated honest-negatives — would be a human confirming that docket posture; the sanction and the quotes are already verifiable from the byte-tied primary and two independent newsroom reproductions.

Verification record
Status
pending
Method
Court-adjudicated. Figures quoted verbatim from the primary court record: the Jan. 2, 2026 Decision and Entry (Doc. #113) and the Feb. 25, 2026 sanction Order (Doc. #122), S.D. Ohio No. 3:25-cv-273, Senior Judge Walter H. Rice. Both PDFs are archive-bound and byte-tied to Wayback (Doc. #122 sha1-b32 AKJALQHV6SANOWMSZS5VJGHKNDINWYQN = CDX digest, capture 20260315185302; Doc. #113 archive-bound from 20260107203121). pdf-parse renders the dollar figures with a font-kerning intra-number space, so the un-spaced $7,500 binds verbatim to two mutually independent archive-bound newsrooms — ABA Journal (via ALM/Law.com) and Mealey's (LexisNexis) — while the byte-tied order supplies the sanction/contempt mechanism and the qualitative findings. Reason/Volokh independently reproduces the Jan. 2 fabrication findings. → two_independent.
Provider
Generative AI, unnamed; Scott admitted using an unspecified generative-AI system 'for preliminary case identification.' The court named no product and said it 'does not know whether' AI generated the caselaw — any 'ChatGPT'/'Claude' attribution would be unsupported
Client
U.S. District Court for the Southern District of Ohio, Western Division at Dayton — Kettering Adventist Healthcare v. Collier et al, No. 3:25-cv-273, Senior Judge Walter H. Rice · Courts / legal (AI-governance honest-negative)
Disclosure
named