The court order arrived in the inboxes of every client Joshua B. Watkins and Burrill Watkins LLC currently represented — not as a private communication between attorneys and their own clients, not as a sealed disciplinary finding, but as a court-ordered instrument of public shaming that a federal judge had commanded be distributed to opposing counsel, to presiding judges across all pending cases, and to every attorney in the Burrill Watkins firm itself.

The document they were forced to send was a detailed federal court opinion cataloging Watkins's AI-related misconduct: fabricated citations, false statements to the court, lies about what he had told his clients, attempts to blame court staff for his own errors, and what the court described as "a course of misconduct" so sustained and willful that it could not be attributed to "ordinary negligence, or even recklessness."

In the history of AI-related attorney sanctions — a history that is now just three years old but already encompasses more than 1,200 documented cases worldwide — no court has ever before ordered an attorney to personally deliver his own disciplinary order to every client he currently represents, as a condition of the sanction itself.

Until now.

What happened to Joshua B. Watkins and his law firm, Burrill Watkins LLC, in a Northern District of Alabama federal court this week is not a story about the hallucination problem. It is a story about what happens when courts stop treating AI sanctions as case-specific remedies and start treating them as instruments of institutional terror — mechanisms designed not merely to correct a particular wrong, but to maximize the professional damage done to any attorney who dares to use the technology without proper reverence for the judiciary's preferred protocols.

The Watkins Case: What Actually Happened

Joshua B. Watkins is an Alabama attorney and partner at Burrill Watkins LLC. He used artificial intelligence to prepare court filings in a federal case. The AI-generated material contained fabricated or inaccurate legal citations — the by now familiar hallucination problem that has afflicted hundreds of attorneys across every jurisdiction in the country since large language models became ubiquitous in late 2022.

What distinguished Watkins's case from the ordinary hallucination incident was what came next: a series of attempts to manage the disclosure, deflect responsibility, and minimize the consequences. Rather than immediately coming clean, accepting responsibility, and doing "the work necessary to establish that it was, in fact, a lapse" — as the court noted it had seen other attorneys do in similar cases — Watkins engaged in what the court characterized as sustained obfuscation spanning months.

He blamed court staff for some of his errors. He buried other instances of misconduct in "convoluted footnotes" within the very filings that were supposed to correct the original problem. He told the court he had kept his clients "fully informed" of the misconduct and the risk to their cases — a claim the court found "unworthy of belief" after hearing from his client, Dulce Rivera, whose family business litigation was at stake. Rivera told the court that Watkins had not advised her about the issues his AI misuse caused. The timing of her decision to ask that Watkins be removed from her family's case — and the timing of his departure from his law firm — contradicted his self-serving narrative at every point.

The court's findings were direct and unsparing: "What began as the misuse of artificial intelligence in a court filing (which Mr. Watkins characterized as an inadvertent technical error) grew into repeated instances of lying and obfuscation." The court found that his conduct "reflects an abdication of Mr. Watkins's duty of candor to the court" and that his false statements were "intentional, repeated choices — not isolated lapses in judgment attendant to unreasonable haste or excusable neglect."

The sanctions imposed were, in their aggregate, extraordinary:

The court publicly reprimanded both Watkins and his law firm. Burrill Watkins LLC — not just the individual attorney, but the entire firm entity — received a formal public reprimand for what the court characterized as an "apparent lack of internal controls and guardrails surrounding its attorneys' use of artificial intelligence — indeed, the very AI the firm pays for and encourages its attorneys to use."

Watkins and the firm were disqualified from further participation in the case. They were ordered to pay the defendants' attorney's fees — a sum that, according to reporting on the order, totals approximately $47,000.

The Clerk of Court was directed to send a copy of the order to the General Counsel of the Alabama State Bar "and any other applicable licensing authorities for further proceedings as appropriate."

And then came the extraordinary measure that sets this case apart from every other AI sanction in the modern era: "Mr. Watkins and Burrill Watkins are ORDERED to provide a copy of this order to their clients, opposing counsel, and presiding judge in every pending state or federal case in which they are counsel of record. They shall also provide a copy of this order to every attorney in their law firm."

Read that again. Every client. Every opposing counsel. Every presiding judge. In every pending case. Every attorney in the firm.

Watkins and Burrill Watkins LLC were not merely sanctioned. They were ordered to personally execute their own professional humiliation — to become the instrument of their own destruction, delivering a detailed record of their misconduct directly into the hands of every person and institution with whom they have a professional relationship.

The Sixth Circuit's Simultaneous Strike: Steven N. Howe

The Watkins order did not arrive in isolation. On April 3, 2026 — four days earlier — the United States Court of Appeals for the Sixth Circuit issued its opinion in United States v. Farris, in which it sanctioned Kentucky attorney Steven N. Howe for AI-assisted brief writing that the court found constituted professional misconduct.

Howe's case was, in some ways, less dramatic than Watkins's. Howe did not fabricate entirely fictitious cases. He used real cases as citations — but the AI he employed had generated quotations from those cases that "do not appear in their cited sources," and the arguments attributed to those cases misrepresented their actual holdings. The fabrications were subtler, harder to detect: not invented authorities but real authorities with invented language.

The Sixth Circuit was unimpressed by the subtlety of the distinction. The court removed Howe from the criminal case — a federal criminal appeal in which his client's liberty was at stake. It denied him all compensation for the work he had performed under the Criminal Justice Act — a financial consequence that, for an appointed defense attorney, amounts to a total forfeiture of income earned. And it referred the matter for potential bar disciplinary action.

The court acknowledged Howe's claim that he had practiced for forty years without prior discipline. It acknowledged, though did not excuse, what it characterized as "stressful circumstances" that may have contributed to the filing of unreviewed AI-generated material. Then it concluded that his failure to verify the citations, regardless of the circumstances, constituted "inexcusable transgressions" that delayed proceedings and "consumed judicial resources."

The Sixth Circuit's ruling included a statement that deserves particular attention: "Attorneys should not utilize technology without knowing the ways in which it can be misused or contribute to inaccuracies." The corollary goes unstated but is clear: until an attorney can demonstrate complete mastery of an AI tool's failure modes, the court's implicit expectation is that the attorney should not use it.

Applied consistently, this standard would prohibit the use of virtually every AI tool currently on the market. It would prohibit the use of Westlaw's AI-assisted research, which has itself generated hallucinated quotations in documented cases. It would prohibit the use of Harvey, CoCounsel, Lexis+AI, and every other tool that the largest law firms in the country have spent hundreds of millions of dollars deploying. It would prohibit the use of any technology that can, under some circumstances, produce inaccurate output — which is, at present, every technology.

The Asymmetry That Reveals the True Agenda

Courts are not held to the same standard they are imposing on attorneys. Approximately 61 percent of federal judges currently use AI in their judicial work — to draft opinions, summarize briefs, identify controlling authority, flag precedent, and manage caseloads. They use the same technology, from some of the same vendors, with many of the same failure modes.

Judges are not required to disclose their AI use. Judges are not required to certify the accuracy of research conducted with AI assistance. Judges face no sanction if an AI tool produces an inaccurate summary of a brief that influences the outcome of a case. Judges do not have to notify anyone when they use AI, and no court has ever ordered a judge to distribute a public reprimand to every party before her because she used an AI tool without perfect mastery of its failure modes.

The asymmetry is not incidental. It is the entire point.

What the Watkins order and the Howe ruling demonstrate, viewed in the context of the now-accelerating pattern of AI sanctions — $109,700 in Oregon, $96,000 in San Diego, mass disqualifications in New Orleans, a separate Alabama case in which a magistrate judge recommended that an attorney be found "incompetent to practice law" — is that federal courts have decided to use the AI hallucination problem as the predicate for a systematic campaign to establish dominance over AI adoption in the legal profession.

This is not a conspiracy theory. It is an institutional behavior pattern with a clear logic: if courts can establish that AI use is inherently risky, that the risk cannot be fully managed, and that the penalties for any AI-related error are catastrophic and career-ending, then the chilling effect on AI adoption will be total. Attorneys who cannot afford to risk their licenses, their clients' confidence, and their entire professional reputation — which is to say, all attorneys — will not use AI.

And if attorneys do not use AI, the profession's economic model remains intact. The billable hour survives. The junior associate survives. The law school debt cycle survives. The access-to-justice gap — which AI could dramatically reduce by making basic legal work fast and cheap enough to serve clients who cannot currently afford attorneys — remains unbridged.

The Extraordinary Remedial Architecture of the Watkins Order

Let us look carefully at what the Watkins court actually ordered, because the architecture of the remedy is revealing.

A typical AI sanction involves: (1) monetary penalty, (2) possible disqualification from the specific case, (3) referral to bar authorities. These remedies are case-specific. They address the harm to the specific court and the specific parties in the specific proceeding. They impose consequences that are proportionate to the wrong committed in that forum.

The Watkins order goes far beyond this. By ordering Watkins to distribute the full text of the sanction order to every client, every opposing counsel, and every presiding judge in every pending case, the court has done something that has no precedent in normal disciplinary practice: it has deputized the sanctioned attorney as an instrument of his own professional destruction, forcing him to personally disseminate the evidence of his misconduct to the maximum possible universe of people with a professional interest in knowing about it.

The court directed the Clerk to publish the order in the Federal Supplement — making it permanently part of the official legal record, findable by any attorney, any law student, any client, or any opposing counsel conducting a Google search of Watkins's name for the rest of his career.

Consider what this order actually accomplishes, functionally:

Every client Watkins currently represents — including clients who hired him for matters entirely unrelated to the AI misconduct — will receive a document describing in detail how their attorney lied to a court, misled his clients, and tried to blame court staff for his errors. Some of those clients will fire him. Some will report him to the bar. Some will demand refunds. Some will retain malpractice counsel.

Every opposing counsel in every pending case will receive the same document. They will immediately use it: in motions, in depositions, in negotiations, in any proceeding where Watkins's credibility is at issue. The order has effectively made Watkins a professionally damaged counterparty in every case he currently handles.

Every presiding judge will receive the document. The next time Watkins appears before any of those judges, every judge will have read, in a federal court opinion, that he has "abdicated his duty of candor to the court" and made "false statements" that were "intentional, repeated choices." The presumption of good faith that attorneys normally receive from the bench will be poisoned for every case Watkins is currently handling.

This is not a remedy designed to correct the wrong done to the parties in the specific case before the court. The parties in that case are made whole by the $47,000 in fee awards and Watkins's disqualification. The mass-distribution order serves a different purpose entirely: it is designed to maximize the professional damage done to Watkins and, more broadly, to send a message to every attorney who reads about the case that AI-related misconduct will not merely end your current case — it will metastasize through your entire practice.

The Firm Sanction: An Unprecedented Institutional Target

Perhaps the most revealing aspect of the Watkins order is its treatment of Burrill Watkins LLC as an independent sanctionable entity. The firm was publicly reprimanded — not because any firm partner or employee other than Watkins was found to have committed misconduct, but because the court found the firm's "apparent lack of internal controls and guardrails surrounding its attorneys' use of artificial intelligence" to be independently sanctionable under Rule 11.

The court's analysis is worth examining. It noted that the firm "has not explained how it enforced any policies about responsible AI use, how it will prevent improper AI use going forward, or any other circumstance — let alone an extraordinary one — why it shouldn't be sanctioned." The firm's argument that it had no reason to know about a partner's AI use was, in the court's view, insufficient: "Though Burrill Watkins's purported lack of knowledge has to do with a partner's misuse of a relatively new technology, its core argument about its own ignorance is both old and ordinary."

Translation: law firms are now on notice that they are responsible not merely for the misconduct of their attorneys, but for the adequacy of their AI governance frameworks. A firm that pays for AI tools and "encourages its attorneys to use" them is, in the court's view, on the hook when those tools are used improperly — regardless of what the firm knew or didn't know.

This creates an almost impossible standard for law firms. If firms prohibit AI use, they risk competitive disadvantage against firms that use it. If they permit AI use, they are potentially liable for any AI-related error committed by any attorney under their roof — even if the firm had policies in place, even if the specific attorney violated those policies, even if the firm acted promptly to remediate when it learned of the problem.

There is no version of this standard under which a firm can both permit AI use and be certain it will not face firm-level sanctions if an attorney misuses it. The court has created a framework in which the safest institutional choice is to prohibit AI entirely — which is, of course, exactly the outcome that would best preserve the economics of traditional legal practice.

The Flood of Sanctions: A Coordinated Pattern

The Watkins and Howe orders arrived in a week in which the pace of AI-related sanctions has accelerated dramatically. Consider the timeline of the past thirty days:

In Oregon, attorney Stephen Brigandi was ordered to pay $109,700 in sanctions — believed to be the largest AI-related penalty in American legal history. In San Diego, Brigandi's law firm faced additional consequences as the court imposed terminating sanctions against the client whose case had generated the hallucinated citations, dismissing her claims with prejudice. In New Orleans, two attorneys resigned from the city's Law Department after being sanctioned for AI-generated fake citations. In Arizona, an attorney was disciplined for AI-fabricated citations in a Phoenix Suns employment discrimination lawsuit. In the Sixth Circuit, the Farris opinion sanctioned Howe and established new precedent for the consequences of unverified AI use in criminal appeals. And in Alabama, the Watkins order established a new category of AI sanction: mass professional humiliation, executed by the attorney himself.

The pace of sanctions in the first week of April 2026 alone — ten cases in ten courts on a single day, according to researcher Damien Charlotin's hallucination database — suggests that the judiciary has moved from reactive punishment of individual violations to something that functions more like a coordinated enforcement campaign.

There is no allegation, and no basis for alleging, that federal judges across multiple circuits and multiple states are coordinating their AI sanction decisions in any improper way. But there is a well-understood dynamic in judicial culture by which high-profile opinions from respected courts are emulated by judges across the system. When a Northern District of Alabama court issues an unprecedented mass-distribution order, that order will be read by federal judges across the country. When the Sixth Circuit removes an attorney from a criminal appeal for AI use in a published opinion, that opinion becomes precedent that other panels can cite.

The result, whether coordinated or not, is a uniform and escalating signal: AI misconduct is a category of professional failure unlike any other, deserving of remedies that are not merely corrective but exemplary — designed not just to punish the individual but to deter every attorney who reads the opinion from making the same choice.

What the Legal Profession Is Actually Losing

The AI hallucination problem is real. Attorneys have submitted fabricated citations to courts. Clients have been harmed by inadequate legal work. Courts have wasted resources investigating and litigating the consequences of AI errors. No responsible observer would argue that courts should simply ignore AI-generated misconduct.

But the remedies being deployed — the mass humiliation orders, the firm-level sanctions, the career-ending removals, the six-figure penalties — are calibrated to achieve something far beyond correcting individual errors. They are calibrated to make AI adoption in the legal profession as costly as possible, as visible as possible, and as terrifying as possible.

The people who will pay the price for this are not the partners at large law firms, who have compliance departments and AI governance frameworks and malpractice insurance that can absorb the occasional sanction. The people who will pay the price are solo practitioners, public defenders, legal aid attorneys, and small-firm lawyers who serve clients who cannot afford Skadden or Latham — the lawyers who most need AI's productivity gains to make legal representation economically viable at the bottom of the market.

When a public defender like Steven Howe cannot use AI to draft a brief without risking removal from the case, denial of compensation, and a referral to bar authorities, the practical effect is that public defenders will avoid AI. And when public defenders avoid AI, their clients — who are, by definition, people who could not afford to hire anyone better — receive slower, less thorough representation than they would if their attorney had access to every available tool.

When a small-firm attorney like Joshua Watkins faces mass distribution of a public reprimand to every client he currently represents, the practical effect is that small-firm attorneys will avoid AI. And when small-firm attorneys avoid AI, their clients — who are, again, people who chose to hire a small firm rather than a large one because they could not afford a large firm — receive services that are more expensive and less efficient than they could otherwise be.

The legal profession's claim to self-governance rests on a bargain with the public: it will use its regulatory power to protect clients, not to protect the profession's economic model. When courts sanction AI use so aggressively that the practical effect is to prohibit the tool that could most dramatically expand access to legal services, the profession has broken that bargain. It has chosen, however unconsciously, institutional preservation over public service.

Conclusion: The Shaming Machine Has a Purpose

Joshua B. Watkins lied to a court. The evidence in the record supports that finding. He deserved to face consequences for that lying. The $47,000 in fee awards, the disqualification from the case, and the referral to the Alabama State Bar are all proportionate responses to documented misconduct that goes beyond mere negligence.

But the mass-distribution order — the requirement that Watkins personally deliver his own court-ordered humiliation to every client, every opposing counsel, and every presiding judge in every pending case — is not about Joshua B. Watkins. It is about every attorney who has considered using AI and is now watching what happened to Joshua B. Watkins.

It is a message, and the message is not subtle: use AI incorrectly, and we will not merely sanction you in the case where you made the mistake. We will make you personally infect every professional relationship you have built with the evidence of your failure. We will make you the instrument of your own professional destruction. We will make the story of your misconduct the first thing every client, every opposing counsel, and every judge knows about you.

Steven N. Howe's situation carries the same message in a different register: use AI in a criminal appeal, and we will remove you from the case, deny you compensation for work you have already done, and send your name to bar disciplinary authorities — regardless of your forty years of prior clean practice, regardless of whether the citations you used actually existed, regardless of whether the error was isolated rather than systemic.

The shaming machine has a purpose. It is not accuracy. It is not proportionality. It is not client protection.

It is preservation — of the economics of traditional legal practice, of the value of a law license as a monopoly credential, of the judiciary's role as the exclusive arbiter of what tools attorneys may use to serve their clients.

The legal profession is not going to win this battle. AI is not going to be frightened away from legal practice by a series of exemplary sanctions, no matter how devastating those sanctions are to individual attorneys. The productivity gains are too large, the competitive pressures too intense, and the client demand for affordable legal services too urgent.

But the sanctions being imposed today are doing real damage along the way. They are doing it to Joshua B. Watkins, whose professional life has been comprehensively unraveled by a sequence of events that began with an unvetted AI draft. They are doing it to Steven N. Howe, who will spend the next chapter of his career explaining a Sixth Circuit opinion to bar investigators and potential clients. They are doing it to every solo practitioner and small-firm attorney who reads these cases and decides that the risk of using AI is simply too high — and whose clients will therefore receive more expensive, slower, less comprehensive legal service as a result.

The courts can call this ethics enforcement. They can call it protecting the integrity of the judicial system. They can call it ensuring that AI is used responsibly in legal practice.

What they cannot call it is what it actually is: a deterrent designed to slow the disruption of an institution that is not ready to be disrupted.

---

Sources and Citations

• Reason.com / Volokh Conspiracy. (Apr. 7, 2026). "AI 'Hallucinated Cases' Lead to $47K Sanctions." reason.com
• U.S. District Court, Northern District of Alabama. Order (Sanction) in Burrill Watkins LLC matter (approximately Apr. 7, 2026). Available via CourtListener
• U.S. Court of Appeals for the Sixth Circuit. United States v. Farris, Opinion (Apr. 3, 2026). Opinion PDF
• Dallas Express. (Apr. 7, 2026). "KY Lawyer Slammed After AI-Generated Briefs Mislead Appeals Court." dallasexpress.com
• Reason.com / Volokh Conspiracy. (Apr. 6, 2026). "$55K Sanctions Related in Part to AI-Hallucination-Filled Court Filings." reason.com
• NPR. (Apr. 3, 2026). "Penalties Stack Up as AI Spreads Through the Legal System." npr.org
• San Diego Union-Tribune. (Apr. 4, 2026). "San Diego Attorney Hit With One of Largest Ever Sanctions for Submitting AI-Hallucinated Filings." sandiegouniontribune.com
• PlatinumIDS Blog. (Apr. 2026). "1,227 Fabricated Citations and Counting: Inside the AI Hallucination Crisis Hitting Courts Worldwide." platinumids.com
• Charlotin, D. (2026). AI Hallucinations in Court Proceedings: Worldwide Tracker. damiencharlotin.com/hallucinations
• Federal Rules of Civil Procedure, Rule 11 (Signing Pleadings, Motions, and Other Papers; Representations to Court; Sanctions).