On May 21, 2026, Judge Harold D. Mooty III of the U.S. District Court for the Northern District of Alabama issued a 34-page sanctions order against attorney H. Gregory Harp. The order publicly reprimanded him, disqualified him from his client's case, suspended him from federal practice in the district for six months, ordered the sanction published in the Federal Supplement, and referred him to the Alabama State Bar and every other jurisdiction in which he is licensed. The triggering conduct was four fabricated case quotations in a brief. The escalation was a $5.27 prorated refund the lawyer requested from OpenAI three days after the court ordered him to produce his ChatGPT history; the refund cut off access before the court's production deadline. The judge's spine sentence reaches every regulated profession in the country.
What actually happened in Miller v. Regions Bank
The case is Jackie L. Miller v. Regions Bank, Case No. 2:24-cv-01324-HDM, pending in the Southern Division of the Northern District of Alabama. Attorney H. Gregory Harp, a solo practitioner, represented Ms. Miller in an Americans with Disabilities Act dispute against the bank.
On November 26, 2025, Attorney Harp purchased a ChatGPT Plus subscription. Four days later, on November 30, 2025, he filed a Response in Opposition to the bank's Motion for Summary Judgment. The brief contained four direct quotations attributed to four real federal cases, complete with case names, reporter citations, and pinpoint page numbers. The court conducted a thorough search for each quotation. None of them existed. The cases were real. The quotations attributed to those cases were fabrications.
The court issued a First Show Cause Order. In his sworn response, Attorney Harp stated that he had used a legal-research database called vLex to write his brief, plus the Microsoft Copilot attachment in Word for grammar and spelling. He did not mention ChatGPT. At a show cause hearing on April 20, 2026, pressed under oath, he admitted he had used ChatGPT "to do a general search," but stated that neither ChatGPT nor any other AI "wrote a single word" of his brief. When asked whether the false quotations were originally generated by ChatGPT, he stated, "Not to my knowledge." Asked whether he had checked his ChatGPT search history to confirm that, he admitted he had not.
Right after the hearing, the court ordered Attorney Harp to produce screenshots of his ChatGPT history for in camera inspection. On April 23, 2026, three days later, an OpenAI email at 6:21 AM informed Attorney Harp that his ChatGPT Plus subscription would not renew but would remain active through April 30. Four minutes later, at 6:25 AM, a follow-up OpenAI message confirmed that he had requested a prorated refund of five dollars and twenty-seven cents. The effect was that access cut off immediately rather than at the end of the billing period.
The court's production deadline was April 29. The original Plus subscription would have remained active through April 30. The $5.27 refund moved the access cutoff to before the deadline. On April 29, Attorney Harp submitted a flash drive containing a one-page unsigned PDF that read, in its entirety: "I have attempted to comply with the Court's Order requiring me to furnish ChatGPT screenshots and native files; however, my account has been deleted, and the files are no longer available."
The prorated refund Attorney Harp requested from OpenAI on April 23, 2026. The original Plus subscription would have remained active through April 30, the day after the court's April 29 production deadline. The refund cut off ChatGPT access before the deadline.
Source: U.S. District Court N.D. Alabama, May 21, 2026 Order, Doc. 39 at 6-7Why the AI was not the problem (the spine sentence)
On page two of the order, the judge places a single paragraph that every executive in every regulated industry should read.
The court is not sanctioning Attorney Harp merely for using generative AI. Attorneys may use such tools responsibly. The sanctionable conduct here includes Attorney Harp's submission of false legal authority, lack of reasonable inquiry, lack of candor to the court, and destruction of evidence during the ensuing sanctions investigation.
The court drew a sharp line. AI use is not misconduct. AI use without verification, without candor, and without preservation of the record is misconduct. A federal judge spent thirty-four pages establishing the distinction.
That distinction is the structural finding worth carrying forward. Every prior high-profile lawyer-AI case, including the 2023 ChatGPT briefs in Mata v. Avianca, has produced commentary in two flavors: AI is dangerous, or lawyers should know better. Neither read is useful. The Alabama order is useful because it names the actual professional standard. Lawyers may use AI, but they must verify load-bearing output, comply with any applicable disclosure rules, and be candid when a court asks how the work was produced. The same logic applies to any regulated work in any industry.
The cover-up arc, in detail
The court's findings of fact build a single picture, sentence by sentence. The picture is not a confused lawyer who panicked. The picture is a deliberate sequence of choices that escalated the harm at each step.
Attorney Harp paid for a ChatGPT Plus account four days before the brief was filed. He used the account in the matter. He filed the brief with four fabricated quotations from real cases. When the court called him on it, he submitted a sworn response that listed vLex and Microsoft Copilot but omitted ChatGPT entirely. At the hearing, under oath, he stated he had used only the free version of ChatGPT. His submitted email screenshots later confirmed he had been on a paid Plus subscription the entire time.
Ordered to produce his ChatGPT history, he canceled the subscription within three days, requested a $5.27 prorated refund, and cut off access before the court's production deadline. He then submitted a one-page unsigned PDF saying the account was gone. The judge wrote: "He has obfuscated the truth at every turn. He has pleaded ignorance as to any wrongdoing, then turned around and destroyed the evidence that likely would have exposed him."
The legal name for this is spoliation of evidence in direct contravention of a court order. The everyday name is the cover-up. The court found that the cover-up, not the original AI use, was the more serious offense.
Lawyers make errors. Competent and ethical lawyers own them.
What "career-altering" actually means
The order ends with a sentence the court appears to have written deliberately for the legal press to quote.
Any lawyer who believes this court will hesitate to vigorously investigate and impose career-altering sanctions for such misconduct is gravely mistaken.
The seven sanctions actually imposed give the phrase concrete shape.
First, a public reprimand on the federal record. Second, a requirement that Attorney Harp personally serve a copy of the sanctions order on every client, every opposing counsel, and every presiding judge in every active state and federal case in which he is counsel of record, within ten days, with a certification of compliance back to the court. For a solo practitioner, that requirement effectively notifies every active professional relationship simultaneously that this lawyer has been publicly reprimanded for AI misuse and evidence destruction.
Third, the order is directed for publication in the Federal Supplement. The Federal Supplement is a widely used reporter for selected federal district court opinions. Publication makes the order easier to find, cite, and discuss in future sanctions analysis, but it does not turn a district court order into binding precedent. The lawyer's name is in the published caption.
Fourth, disqualification from this case. The plaintiff has sixty days to associate new counsel. Fifth, a six-month suspension from practice in the United States District Court for the Northern District of Alabama, effective fourteen days from the entry of the order. Sixth, Attorney Harp must provide the court within twenty-four hours a listing of every jurisdiction in which he is licensed to practice. Seventh, the Clerk of Court is directed to serve a copy of the sanctions order on the General Counsel of the Alabama State Bar and every other applicable licensing authority for further proceedings as appropriate.
The combination is what makes the phrase load-bearing. A six-month federal suspension can be served. But a federal sanctions order in the Federal Supplement, plus a referral to every state bar in which a lawyer is licensed, plus a personal notification requirement to every active client and opposing counsel, could be professionally devastating for a solo practice and a permanent mark on every regulated professional record the lawyer holds.
That is what the judge meant by career-altering.
Where this maps in The 7 Levels of AI Proficiency
The Alabama order reads like a proficiency framework written by a federal judge. The court is naming behaviors and assigning consequences based on the distance between the behavior and the professional standard. The 7 Levels of AI Proficiency framework names those same behaviors and assigns measurable progression between them.
At Level 1, the Cadet, someone is AI Aware. They know AI exists and have basic awareness of its applications. At Level 2, the Ensign, someone has begun using AI tools but has not yet built reliable verification habits. They produce work product that may include AI output, and they have not yet developed a practice for checking it. At Level 3, the Lieutenant, someone is a Critical Thinker about AI. They verify before relying. They disclose when they have used AI in load-bearing ways. They are aware of where AI tools fail and apply judgment accordingly. At Level 4, the Commander, someone is a Context Engineer. They integrate AI into structured workflows with verification gates built in. They build organizational practices that allow others to use AI safely. They understand governance posture.
Attorney Harp, on the public record of the court order, was operating at Level 2. He used a paid AI tool in a load-bearing way (preparing a federal brief). He did not verify the output (the four fabricated quotations went through unchecked). He did not disclose the use accurately when first asked (he omitted ChatGPT, then claimed only the free version, then minimized it as "a general search"). He did not maintain a workflow record (the deletion). His professional role required Level 4 or above. The duty of competence under American Bar Association Model Rule 1.1, including its technology comment, requires that a lawyer using a technology must understand its capabilities and limitations. The duty of candor under Rule 3.3 requires accurate representation to the tribunal. The Rule 11 inquiry obligation requires reasonable investigation of the legal authority cited.
The distance between Level 2 and Level 4 is the distance the court named when it described the misconduct. The framework is what lets an organization measure that distance before a court does.
This is also why a blanket "ban AI" policy does not solve the problem. The court explicitly said attorneys may use AI tools responsibly. A ban does not raise proficiency. A ban only delays the moment when someone uses AI without permission, without verification, and without the workflow to catch the error. The fix is proficiency, not prohibition.
Why this applies to every regulated profession, not just law
The legal-AI hallucination story has been told as a profession-specific narrative since 2023. The Alabama order is the moment the structural finding becomes portable to every other regulated profession.
Every regulated profession has a duty of competence. The American Institute of Certified Public Accountants requires due professional care. State medical boards require the standard of care. State engineering boards require professional engineering competence. The Financial Industry Regulatory Authority requires reasonable-basis suitability. The Securities and Exchange Commission requires investment-adviser fiduciary care. Those standards already point in the same direction: professionals remain responsible for the work product they submit, even when AI helped produce it. Disclosure duties vary by profession, rulebook, and context, but verification responsibility does not disappear because a tool was involved.
Every regulated profession also has a duty of candor in its own form. Accountants have it in audit opinions. Doctors have it in informed consent. Engineers have it in stamped drawings. Financial advisors have it in disclosure documents. The Alabama order's reasoning maps onto each of them. A regulated professional submitting work product without verifying AI-generated content has breached the duty of competence. A regulated professional concealing the use, then destroying evidence of the use, has breached the duty of candor.
The Alabama State Bar referral is the procedural model every other regulated body has the authority to follow. A state CPA board reviewing a complaint can request the underlying work papers, including AI tool logs. A medical board reviewing a complaint can request the diagnostic decision audit trail. The destruction-of-evidence overlay in the Alabama order is what every regulator will read carefully, because it tells them how to structure their next complaint procedure.
For Indiana specifically, the connection is direct. Indiana Rule of Professional Conduct 1.1 requires competent representation and includes keeping up with the benefits and risks of technology relevant to a lawyer's practice. Indiana Rule 3.3 requires candor toward the tribunal and bars knowingly false statements of fact or law. Indiana Trial Rule 11 also serves a certification function for signed pleadings and motions, though it is not identical to Federal Rule 11. An Indiana lawyer engaging in the same fact pattern would face serious sanctions and disciplinary exposure under closely related professional standards. Indiana's other regulated professions sit under analogous frameworks.
Specific sanctions imposed in one order: public reprimand, mandatory notification to every active client and opposing counsel, publication in the Federal Supplement, disqualification, six-month federal practice suspension, jurisdiction listing requirement, and referral to the state bar plus every other licensing authority.
Source: U.S. District Court N.D. Alabama, May 21, 2026 Order, Doc. 39 at 33-34Three things to do with this order this week
The Alabama order is a credible, federal, published signal about how the next decade of AI accountability will work in regulated professions. Here is how to use it practically.
Inventory who is using AI in regulated or fiduciary work.
For every role in your organization that produces work product covered by a duty of competence (legal briefs, audit opinions, engineering calculations, medical decisions, financial advice, board reports), surface who is using AI today and how. If you cannot answer that question yet, the inventory is the first build. A policy document without an underlying inventory is a sentence, not a posture.
Establish a verification and disclosure protocol that meets your industry standard.
What does the relevant duty of competence require when AI is used in load-bearing work? What disclosure does your industry rulebook require? Write the protocol against those specific standards, not against a generic AI policy template. The standards are already there. The Alabama order is the warning shot that they will be enforced.
Measure the proficiency shortfall before a court or regulator measures it for you.
Run a structured AI proficiency assessment across your team. The 7 Levels of AI Proficiency assessment at assess.launchready.ai places each person on a measurable scale in about ten minutes. Use the results to identify the distance between where each person sits and where their role actually requires them to be. The distance is the work. Close it through training and process before someone has to close it through a sanctions hearing.
Related reading: The 7 Domains of AI Governance: A Framework for Mid-Market Leaders and Level 3: The Lieutenant in the 7 Levels of AI Proficiency.
Sources
- Miller v. Regions Bank, Case No. 2:24-cv-01324-HDM, Order (Doc. 39), U.S. District Court for the Northern District of Alabama, Southern Division, May 21, 2026.
- Reuters. "Judge says lawyers' AI use risks career-altering consequences." May 22, 2026.
- American Bar Association. Model Rules of Professional Conduct (Rule 1.1 Competence, Rule 3.3 Candor Toward the Tribunal).
- Federal Rules of Civil Procedure, Rule 11 (Signing Pleadings, Motions, and Other Papers; Representations to the Court; Sanctions). Legal Information Institute, Cornell Law School.
- Painter, Harrison. "The 7 Domains of AI Governance: A Framework for Mid-Market Leaders." LaunchReady.ai Insights.
- The 7 Levels of AI Proficiency framework. LaunchReady.ai.
Frequently Asked Questions
What did the Alabama federal court order on May 21, 2026?
Judge Harold D. Mooty III of the U.S. District Court for the Northern District of Alabama issued a 34-page sanctions order against attorney H. Gregory Harp in Miller v. Regions Bank (Case No. 2:24-cv-01324-HDM). The order publicly reprimanded Harp, disqualified him from the case, suspended him from practice in the Northern District of Alabama for six months, ordered the sanction published in the Federal Supplement, required him to share the order with every client and opposing counsel in every active case, and referred the matter to the Alabama State Bar and every other jurisdiction in which he is licensed.
Was the lawyer sanctioned for using AI?
No. The court was explicit: "The court is not sanctioning Attorney Harp merely for using generative AI. Attorneys may use such tools responsibly." The sanctionable conduct was submitting false legal authority (four fabricated case quotations) without reasonable inquiry, lack of candor to the court, and the destruction of evidence during the sanctions investigation. The AI itself was not the offense.
What did the lawyer actually do?
Attorney Harp purchased a ChatGPT Plus subscription on November 26, 2025, then filed a Response in Opposition four days later containing four fabricated quotations from real cases. When confronted, he initially claimed he did not use AI to generate the quotations and that he used only the free version of ChatGPT. Three days after the court ordered him to produce his ChatGPT history on April 20, 2026, he canceled his Plus subscription on April 23 and requested a prorated refund of $5.27 to lose access immediately rather than at the end of his billing cycle on April 30, which was after the court's production deadline.
What does "career-altering" mean in the order?
Judge Mooty wrote: "Any lawyer who believes this court will hesitate to vigorously investigate and impose career-altering sanctions for such misconduct is gravely mistaken." The phrase is not metaphor. A six-month suspension can be served. A federal sanctions order published in the Federal Supplement and served on every state bar in which a lawyer is licensed could be professionally devastating for a solo practice and a permanent mark on a regulated professional record.
Does this case apply to professions outside law?
The pattern applies to any regulated profession with a duty of competence and a duty of candor: medicine (state medical boards), accounting (AICPA and state CPA boards), engineering (state PE licensure), financial advice (FINRA and SEC rules), and others. The reasoning in the order is structural. When a regulated professional uses AI to produce work product, the duty of competence requires verification and the duty of candor requires disclosure. The Alabama State Bar referral pattern is the model every other licensing body has the authority to follow.
What does this have to do with The 7 Levels of AI Proficiency?
The court's read maps directly to the framework. The attorney was operating at Level 2 of The 7 Levels of AI Proficiency (someone who uses AI tools without consistent verification or disclosure) in a profession that requires Level 4 or above (someone who integrates AI with verification workflows, governance posture, and full candor about how the work was produced). Professional consequences attach when the proficiency distance is large and the work is regulated. The framework gives an organization a way to measure that distance before a court does.
What should a CEO or general counsel do this week?
Three actions. First, inventory who in the organization is using AI in regulated or fiduciary work, and assess what level of AI proficiency they actually have. Second, establish a verification and disclosure protocol that meets the relevant industry duty of competence. Third, run a structured assessment across the team to surface the distance between current proficiency and required proficiency, then close it through training and process. The Alabama order is the early warning that this is no longer a discretionary build.
How does this apply to Indiana lawyers and Indiana businesses?
Indiana Rule of Professional Conduct 1.1 requires competent representation and includes keeping up with the benefits and risks of technology relevant to a lawyer's practice. Indiana Rule 3.3 requires candor toward the tribunal and bars knowingly false statements of fact or law. Indiana Trial Rule 11 also serves a certification function for signed pleadings and motions, though it is not identical to Federal Rule 11. An Indiana lawyer engaging in the same fact pattern would face serious sanctions and disciplinary exposure under closely related professional standards. Indiana businesses with in-house counsel, outside counsel relationships, or any regulated professional staff need to read the Alabama order as a forward-looking notice of how state disciplinary processes will treat AI misuse and cover-up.
Find your AI Proficiency level
The free 7 Levels assessment places you across seven stages of AI capability. Under ten minutes. Research-backed scoring.