By Chuck Gallagher — Business Ethics Keynote Speaker and Trainer
TL;DR: The Oregon Supreme Court issued its first-ever sanctions orders against litigants who submitted filings containing fabricated case citations and invented quotes generated by an AI tool called “LegalAI” — striking one petition entirely and dismissing the proceeding. Chuck Gallagher, business ethics keynote speaker and AI speaker and author, argues that this is not a technology failure but an ethics failure: a predictable outcome when speed displaces accountability and no one owns the obligation to verify truth before it reaches a court of law.
Imagine filing a legal brief with the Oregon Supreme Court — a document that cites cases that do not exist, quotes judges who never said those words, and attributes the whole fabrication to an AI tool. Now imagine doing it twice. The second time within twelve hours of being told by the court to stop.
That is what happened in Aldridge v. Tussing (S072780), one of two cases in which the Oregon Supreme Court issued its first-ever orders sanctioning litigants for AI-generated fabrications in their filings. In a separate matter, Witkin v. McGreevy (S072692), the court allowed an amended filing only after a $500 monetary sanction — the respondent had admitted that AI produced “fictitious authorities and inaccurate legal arguments” in the original submission.
The court’s language in the Aldridge matter was direct: “The injection of false precedent undermines the integrity of the proceedings. Doing it repeatedly, and in response to a show cause order, warrants imposition of a meaningful sanction.” The entire proceeding was dismissed. Not delayed. Dismissed.
What Does “Fabricated” Actually Mean Here?
Some people read stories like this and assume a minor citation error — a wrong page number, a misquoted statute. That is not what this is. The litigants in Aldridge cited cases that do not exist. They included quotes attributed to judges who never wrote them. The citations were inventions, and “LegalAI” — the generative AI service they relied on — produced them with the same confident formatting it would use for a real case. The tool did not flag uncertainty. The litigants did not verify. The brief landed on the court’s desk as if it were grounded in real law.
As a business ethics keynote speaker and AI speaker and author, I have written extensively at ChuckGallagher.com about the pattern that shows up in nearly every institutional ethics failure I have studied: need, opportunity, and rationalization. The litigants had a need — to build the strongest possible case, quickly. The AI tool provided the opportunity to generate authoritative-sounding citations without doing the hard work of actual legal research. And the rationalization wrote itself: “It’s research. It’s faster. We’ll sanity-check it later.” Except “later” arrived after the filing was already in the court record. The rationalization collapsed the moment it met reality.
What makes this case especially striking is the sequence of events after the first order. The relators acknowledged the fabrications. They acknowledged the strain on the judicial system. And then, less than twelve hours later, filed another document containing more fabrications. That is not carelessness. That is an organizational process with no internal check — no one whose job it is to confirm that a citation is real before it leaves the building.
Is This an Isolated Incident — or the Beginning of a Trend?
The Oregon Judicial Department’s press release is careful to note that the Oregon Supreme Court and the Court of Appeals are experiencing “a noticeable increase in the number of filings containing such fabrications attributed to AI.” Court of Appeals Chief Judge Erin C. Lagesen wrote publicly on April 29, 2026 about “a rapidly escalating number of filings that contained fabricated legal authority likely produced by the use of generative artificial intelligence,” adding that these filings were being submitted by lawyers and self-represented litigants alike, and that they “create a significant drain on court resources that, at present, appears to be growing.”
This is not an Oregon problem. Federal courts across the country have issued similar sanctions, most visibly in the 2023 case of Mata v. Avianca in the Southern District of New York, where two attorneys were sanctioned after submitting a brief that cited six AI-generated, nonexistent cases. The attorneys had used ChatGPT and did not verify the citations. The judge in that case called the conduct “unprecedented.” Two years later, Oregon is learning the same lesson — and the volume appears to be increasing, not decreasing.
Chief Justice Meagan A. Flynn put it plainly: “When the court has to spend considerable time and effort addressing fabricated legal arguments, it comes at the expense of other cases.” That sentence is about resources. But underneath it is something more fundamental. Courts run on trust. Every party in a proceeding — the opposing counsel, the judge, the appellate panel — assumes that citations are real. The moment that assumption has to be verified for every filing, the whole system slows down and the cost falls on everyone except the person who filed the fabrication.
What Leaders Can Actually Do About This
I want to be direct: the problem here is not AI. AI did exactly what it tends to do — produce confident-sounding output whether or not that output is grounded in fact. The problem is that no one made citation verification a required, accountable step before the filing was submitted. When verification is not someone’s assigned responsibility, it becomes everyone’s assumption. And assumptions are where institutional ethics failures are born.
As an AI ethics speaker and author, I frame this as a governance question, not a technology question. Any organization using AI to assist in work product that goes outside the organization — whether it is a legal brief, a clinical report, a federal contract, or a public statement — needs a documented, human-verified checkpoint before that work product leaves the building. This is not about distrust of AI. It is about assigning clear accountability for accuracy. If your workflow does not require a human being to confirm that every citation, every statistic, every quotation is real before submission, you have an internal control gap. And internal control gaps are where the Oregon Supreme Court’s docket fills up.
The NIST AI Risk Management Framework identifies “Govern,” “Measure,” and “Manage” as core functions for responsible AI use — which in practice means defining who owns accountability for accuracy, building testing mechanisms to catch AI errors before they cause harm, and monitoring for accuracy failures in real time. That is not aspirational policy language. That is exactly the internal control that was missing from the Aldridge filing process. If someone had been required to sign off on every citation before submission, the first set of fabrications would have been caught before reaching the court. The proceeding would still be alive.
Frequently Asked Questions
What did the Oregon Supreme Court sanction litigants for in its first AI-related orders?
The Oregon Supreme Court sanctioned litigants in two separate cases for submitting filings that contained fabricated case citations and invented quotes produced by a generative AI service. In Aldridge v. Tussing (S072780), the court struck the petition and dismissed the proceeding entirely. In Witkin v. McGreevy (S072692), it allowed an amended filing after a $500 sanction. The court called the injection of false precedent an undermining of the integrity of the proceedings.
What is “LegalAI” and how did it produce fabricated citations?
LegalAI is the name of the generative AI service the litigants in Aldridge v. Tussing reported using to conduct their legal research. Like other large language model tools, LegalAI can produce case citations and legal quotations that appear credible in format but are entirely invented — a well-documented phenomenon in AI known as “hallucination.” The litigants indicated they had also relied partly on Google Scholar in their review, but the fabrications were not caught before filing.
Are AI citation errors in legal filings a widespread problem?
Yes, and the frequency appears to be growing. Court of Appeals Chief Judge Erin C. Lagesen wrote publicly in April 2026 that Oregon courts were receiving a “rapidly escalating number” of filings with fabricated legal authority attributed to AI, submitted by both attorneys and self-represented litigants. The phenomenon gained national attention after the 2023 Mata v. Avianca case in the Southern District of New York, where two attorneys were sanctioned for submitting a brief citing six nonexistent AI-generated cases.
What should organizations do to prevent AI-generated errors in legal or formal filings?
Chuck Gallagher, AI ethics speaker and author, recommends treating AI-assisted work product as a controlled process: require a named human being to verify every citation, statistic, and quotation before any document is submitted outside the organization. This aligns with the NIST AI Risk Management Framework’s “Govern” and “Manage” functions, which call for explicit accountability structures and monitoring for accuracy failures. The root cause in the Oregon cases was not a technology error — it was the absence of an internal checkpoint that made verification someone’s defined responsibility.
Does the Oregon Supreme Court’s ruling apply only to attorneys or also to people representing themselves?
The Oregon Judicial Department’s statement makes explicit that the obligation to submit accurate information extends to individuals representing themselves — called pro se litigants — and not only to licensed attorneys. This is significant because many of the AI-assisted fabrication cases in courts nationally have involved self-represented individuals who may not fully understand that AI-generated legal research must be independently verified before use in any court filing.
Join the Conversation
The Oregon Supreme Court just delivered a lesson that every organization using AI needs to hear — and the lesson is not about AI. It is about accountability. When speed becomes the primary value and verification becomes optional, fabrications do not stay inside the tool. They find their way into briefs, reports, contracts, and public statements. I want to know: what internal checkpoints does your organization have in place before AI-assisted work leaves the building? Share your thoughts in the comments below — I read every one of them. That question naturally leads to the questions below, which I hope will prompt deeper reflection on what accountability looks like in an AI-enabled workplace.
Five Questions for Further Thought and Consideration
- If AI-generated content is submitted under your name — whether in a legal brief, a report, or a client proposal — who in your organization is currently accountable for verifying its accuracy before it leaves the building?
- The litigants in the Aldridge case acknowledged the problem to the court and then filed another fabricated document twelve hours later. What does that sequence tell us about the difference between acknowledging an ethical failure and actually fixing the process that caused it?
- Verification requirements are a form of internal control. Where else in your organization have you removed internal controls in the name of speed — and what is the realistic worst-case consequence if an error reaches an external audience?
- Chief Justice Flynn noted that the court’s time spent on fabricated filings comes at the expense of other cases. When one party’s ethical shortcut imposes costs on an entire system, how should those costs be assigned — and does a $500 sanction match the magnitude of the harm?
- The fraud triangle — need, opportunity, and rationalization — maps cleanly onto these AI citation cases. What structures in your professional environment are currently reducing each of those three forces, and which one is receiving the least attention?
