State National v. Treadwell: Public Reprimand for ChatGPT and OpenCase Hallucinations | Advocate Prakhar

Case at a Glance

Full Citation	State National Insurance Co. v. Treadwell, Case No. 2:24-cv-1424-HDM (N.D. Ala. March 27, 2026)
Court	U.S. District Court, Northern District of Alabama, Southern Division
Date	March 27, 2026 (show-cause hearing March 26, 2026)
Judge	Harold D. Mooty III, U.S. District Judge
Category	AI Hallucination — Multi-AI Workflow; Public Reprimand
Jurisdiction	USA — Federal (Alabama)
AI Tools	ChatGPT and OpenCase (both explicitly named)
Outcome	Public reprimand; order to be circulated to all opposing counsel and presiding judges in pending cases; published in Federal Supplement; no monetary fine in this instance

Background

State National Insurance Company sued Damon and Catherine Treadwell seeking final judgment in a coverage dispute. Attorney Edward Eugene May II, representing the Treadwells, filed a timely response containing two purported quotations from two real federal cases: one from the Fifth Circuit’s Nishimatsu Construction decision (1975) and one from the D.C. Circuit’s Whelan v. Abell (1992). Judge Mooty’s chambers research revealed that neither quotation appeared anywhere in the cited cases, and after further research the court concluded that “neither quote exists at all.” May was ordered to show cause.

The AI Issue

May’s response disclosed a multi-step AI drafting process: he uploaded source materials into ChatGPT to produce a preliminary outline, then uploaded that outline into OpenCase to produce a refined draft, repeated the cycle until satisfied, copied the result into a word processor, and edited it into a final version. He admitted that while editing he did not “thoroughly check all citations to ensure that each quote included in the final version did, in fact, exist.” May confirmed the quotations were hallucinations generated by one or both AI tools. The court found May’s conduct “akin to contempt” — the standard for sua sponte Rule 11 sanctions — and noted aggravating factors: his response brief was filed twenty-six minutes late, contained typographical errors (“Courts [sic] Order to Show Casue [sic]”), and he arrived approximately ten minutes late to the show-cause hearing.

What the Court Decided

• The multi-step AI drafting process does not transfer citation-verification responsibility to the software — the attorney’s duty remains nondelegable. [personal responsibility doctrine]
• Fake legal citations “pose a serious threat to the fair administration of justice” requiring equally serious sanctions. [systemic harm doctrine]
• Courts cannot rely on specific cited authority if that authority is fabricated — even if the attorney brings the correct legal issues to the court’s attention. [accuracy over substance]
• May was publicly reprimanded — the reprimand order directed for publication in the Federal Supplement. [public reprimand as deterrent]
• May was ordered to provide a copy of the reprimand order to opposing counsel and the presiding judge in every pending state or federal case in which he is counsel of record. [systemic notification sanction]
• No monetary fine, no disqualification, no bar referral — but all options expressly reserved for future violations. [escalating consequences warning]

“Every second the court spends doing a lawyer’s job for him — as this court was compelled to do for Mr. May — is a disservice to his clients and a waste of taxpayer dollars that needlessly defers other necessary work.”

— Judge Harold D. Mooty III, N.D. Alabama, March 27, 2026

The India Angle

Indian Law Equivalent

India’s federal equivalent of the N.D. Alabama’s public reprimand is a judicial censure published in the court’s order, which becomes part of the permanent record. Under Rule 11 of the Civil Rules and Section 35 of the Advocates Act 1961, attorneys who engage in conduct unworthy of a legal practitioner can be suspended or disbarred. The Bar Council of India’s Disciplinary Committee routinely receives complaints involving professional misconduct — AI-hallucinated citations would clearly fall within this category as it matures into Indian practice.

Bar Council Rules

Rule 17 of the BCI Rules requires advocates to conduct litigation with dignity and in a manner compatible with the dignity of the judicial office. A public reprimand for AI hallucination would significantly damage an advocate’s professional standing before the courts in which they practice — particularly in smaller bar communities where reputation is a core professional asset.

Practical Advice for Indian Advocates

• Never use a “pipeline” AI drafting approach — generating content in one AI tool and refining in another — without a complete citation audit at the final output stage.
• Build a checklist habit: before filing any court document, physically verify every quoted language against the source judgment, ticking off each verified citation with initials and date.
• Punctuality and professional presentation at show-cause hearings matter — the Treadwell court documented May’s late filing and tardiness to the hearing as aggravating factors that could influence the severity of sanction in future incidents.

Quick Takeaways

• Using multiple AI tools in sequence does not reduce the attorney’s verification responsibility.
• Public reprimand orders published in legal databases are a powerful deterrent — they follow an attorney’s professional record indefinitely.
• Lateness, typographical errors, and tardiness at sanctions hearings compound the original AI hallucination violation.

---

Deep Dive: Multi-AI Workflows, OpenCase, and the Verification Gap

State National v. Treadwell introduces a phenomenon that courts are only beginning to grapple with: the multi-AI drafting workflow. Attorney May did not simply paste a prompt into ChatGPT and file the result. He engaged in a sophisticated, iterative process — using ChatGPT for initial structure, OpenCase for refinement, and a word processor for final editing. This level of process sophistication creates a cognitive illusion of thoroughness: the attorney has handled the document many times, revised it substantially, and produced what feels like a carefully crafted final product. Yet the citation verification step was absent from every stage of the pipeline.

This case reveals a critical gap in how legal AI tools are marketed versus how they are actually used. OpenCase, like other legal AI platforms, is positioned as a tool that integrates legal research with drafting — the implicit promise being that citations generated in the drafting process are grounded in real legal authority. But the Treadwell case demonstrates that even purpose-built legal AI tools can hallucinate plausible-sounding quotations for real cases. When an attorney uses two AI tools in sequence, each can inherit and amplify the other’s errors, producing a document that has been “refined” but not verified.

Judge Mooty’s decision to order publication of the reprimand in the Federal Supplement — a standard legal reporting service — rather than simply issuing an internal court order reflects a broader policy purpose: deterrence by publicity. The Treadwell reprimand order will appear in legal databases that attorneys, law firms, and clients use for due diligence searches. Future clients of May who run a routine check will find the reprimand. Law firms that cross-reference opposing counsel will find it. This reputational dimension of public sanctions is qualitatively different from a monetary fine, which can be absorbed as a cost of doing business.

For Indian advocates who are increasingly using AI-assisted drafting in insurance, civil, and commercial litigation, the Treadwell case is a masterclass in what not to do. The state of Indian insurance litigation — particularly third-party motor accident claims under the Motor Vehicles Act 1988 — involves high-volume filings where AI tools offer genuine efficiency gains. But the verification discipline must match the drafting speed. A verification checklist built into every filing workflow is not merely best practice in 2026; it is the minimum standard that Indian courts, following global precedent, will increasingly expect.