TL;DR: A federal judge sanctioned attorney Jocelyn Stewart $3,000 on July 24, 2026 after her filings cited a Washington Supreme Court case that does not exist, calling the conduct tantamount to bad faith. She is one of at least five people sanctioned in 2026 for unverified AI output in court filings, totaling roughly $145,000 in Q1 alone. The failure is never the AI tool, it is the missing verification step before the output left the building. Below is the 2026 tracker plus a copy-paste AI output verification policy any small team can adopt this week.
The case that just happened
On Friday, July 24, 2026, US District Judge Tiffany M. Cartwright of the Western District of Washington sanctioned attorney Jocelyn C. Stewart $3,000 in LeDoux v. Outliers, Inc., a product liability case over nootropic supplements that a nurse alleged contained undisclosed amphetamines and cost her a military drug screening.
The sanction was not about the underlying case. It was about how Stewart argued it.
Her filings contained dozens of inaccurate factual and legal citations spread across at least five separate documents. One of the fabricated authorities was cited as a Washington Supreme Court case, Berg v. Chevrolet Motor Div., 84 Wn.2d 102. That case does not exist. When opposing counsel and the court flagged the problems, Judge Cartwright ordered Stewart to show cause why she should not be sanctioned. Rather than acknowledging the underlying citations were unsupported, Stewart's subsequent corrections introduced new false quotations and misstated what the real, cited authority actually said.
That second failure is what pushed the judge from a warning to a sanction. Cartwright's order described the conduct as tantamount to bad faith, not because one AI tool hallucinated once, but because the correction process failed to catch it too.
This is not a story about a careless lawyer. It is a story about a missing verification step, repeated twice.
This was not an isolated case
Stewart's sanction is the most recent entry in a growing 2026 pattern. Courts have been fining attorneys and referring them for discipline all year over AI-generated content nobody checked before it was filed.
| Case | Court | Sanction | Date | What happened |
|---|---|---|---|---|
| LeDoux v. Outliers, Inc. | W.D. Wash. (3:24-cv-05808) | $3,000 | Jul 24, 2026 | Filings cited a nonexistent Washington Supreme Court case; corrections introduced more false material instead of fixing it |
| Whiting v. City of Athens | 6th Cir. (No. 25-5425) | $15,000 each for two attorneys, plus opposing fees and double costs | Mar 13, 2026 | Appellate briefs contained more than two dozen citations that were incorrect, misrepresented, or entirely nonexistent |
| Oregon Court of Appeals brief | Or. Ct. App. | $10,000 | Mar 2026 | An opening brief contained at least 15 false citations and 9 quotations the court said were contrived from thin air |
| Couvrette v. Wisnovsky | D. Or. | $15,500 plus adverse costs | Q1 2026 | Three summary judgment briefs in a family winery dispute contained 15 fabricated citations and 8 false quotations |
| United States v. McGee | S.D. Ala. | $5,000 | Oct 14, 2025 | Bogus citations in a criminal case drew a referral for removal from court-appointed case eligibility |
Court-tracked sanctions for fabricated AI citations totaled roughly $145,000 in the first quarter of 2026 alone, according to legal-industry tracking published by ComplexDiscovery. That figure covers only the cases that became public through published orders. It says nothing about the filings, reports, and client memos that contained the same kind of unverified AI output but never reached a judge who checked the citations.
Every case on this list shares the same root cause. Nobody actually read the source material before it went out the door.
Why this is not just a law firm problem
It is tempting to read this as a legal industry story. Lawyers have a professional duty to verify citations, a court record makes their failures public, and judges write opinions that name names. None of that is true for most other functions inside a small business.
That does not mean the exposure is smaller. It means it is invisible until it is not.
A marketing team that lets an AI tool draft a claim about a competitor's product, a study that does not say what the AI said it says, or a statistic pulled from a source that was never checked, is one FTC complaint away from the same kind of sanction, just filed under deceptive advertising instead of Rule 11. This site has already covered how the FTC treats AI-generated marketing claims the same way courts treat AI-generated citations: unverified is the violation, regardless of intent.
A grant writer, a compliance analyst preparing a regulatory filing, a customer success rep drafting a client-facing incident report, an account exec building a sales deck with AI-generated case studies. Every one of those roles can produce content that leaves the organization carrying a fabricated fact nobody checked. The difference between that and Jocelyn Stewart's filing is only that a judge has not read it yet.
Small teams are more exposed than large ones here, not less. A large firm has a research department, a paralegal pool, or a compliance review layer that catches some of this by accident, through sheer redundancy. A five-person team asking an AI tool to draft a client report and shipping it the same afternoon has no such safety net. The verification step has to be designed in on purpose, because nothing else is going to catch the mistake first.
The actual failure, every time
Look at the tracker again. None of these sanctions happened because an AI tool hallucinated. Every generative AI tool available in 2026 will occasionally invent a citation, a quote, or a fact that sounds plausible and is not real. That is a known, well-documented limitation, not a surprise.
The sanctions happened because nobody checked the output against a primary source before it left the building. In Stewart's case, the failure happened twice: once when the original filings went out unchecked, and again when the corrections, meant to fix the problem, were also not checked against the actual case law.
This is the part every governance conversation about AI tends to skip. The policy question is not "should we allow AI tools." Every team on this list was already using AI tools; that decision was made and is not coming back. The policy question is "what happens to AI output between the moment it is generated and the moment it reaches someone outside the organization." For every sanctioned case above, the honest answer was: nothing.
A copy-paste AI output verification policy
Here is a short policy any small team, legal or otherwise, can adopt this week. Paste it into your AI acceptable use policy or your client-deliverable checklist as-is, and adjust the named roles to match your team.
AI OUTPUT VERIFICATION POLICY
1. Any AI-generated citation, statistic, quote, case reference, or named
source must be independently verified against its primary source before
it appears in anything that leaves the organization: filings, client
deliverables, marketing claims, regulatory submissions, or public
statements.
2. "Independently verified" means a named human opened the primary source
itself, not another AI summary of it, and confirmed the fact matches.
3. Every external-facing document produced with AI assistance must carry a
verification log: who checked it, what was checked, and the date.
4. AI-drafted corrections or revisions require the same verification as the
original draft. A correction is not exempt just because it fixes a
flagged error; it can introduce new ones.
5. No AI-generated content moves from internal draft to external delivery
without passing through step 1. This applies regardless of deadline
pressure.
6. Any AI tool used to produce external-facing content must be listed in
the team's approved-tools register. Unlisted tools require sign-off
before use on anything client-facing.
That is the entire fix. It costs nothing beyond the time it takes to read a source document, and every sanction on the tracker above happened at a team that skipped exactly this step.
Five questions to ask before anything AI-drafted goes external
- Did a human open the primary source, or just read the AI's summary of it? These are not the same activity. Reading a summary is how the Sixth Circuit case happened.
- Does the verification log show who checked this, and when? If the answer is "we're pretty sure someone looked at it," that is a no.
- Was the correction checked as carefully as the original? Stewart's case shows the correction can be where the real damage happens.
- Is the AI tool that produced this content on your approved list? If nobody knows which tool drafted a given paragraph, nobody can flag the tools with worse hallucination rates.
- Would this survive an opposing party, a regulator, or a client running the same citations through a search engine? That is effectively what happened in every case on the tracker above.
None of these questions require a security team, a legal department, or a budget. They require someone with the authority to say "we are not sending this yet."
What changes for small teams starting now
The 2026 sanctions tracker is going to keep growing. AI adoption for drafting, research, and first-pass content generation is not slowing down at small companies, and the base rate of hallucination in generative AI tools has not dropped to zero. What has changed is that courts, and increasingly regulators, have stopped treating a fabricated citation as an honest mistake deserving a warning. Judge Cartwright's phrase, tantamount to bad faith, is a signal about where the tolerance line now sits.
The fix does not require abandoning AI tools. It requires treating AI output the way a careful editor treats a first draft from a junior researcher: useful, often accurate, and never trusted without a check against the original source. Teams that build that check into their workflow now will not be the next row on this tracker. Teams that keep shipping AI output on faith will be.
Related Reading
- AI Acceptable Use Policy: Annual Update Guide 2026
- TypeScript AI Agent Output Validation Patterns 2026
- Vetting AI Tools: Fake Apps and Malware 2026
- FTC AI Marketing Claims Checklist 2026
- AI Vendor Contract Red Flags 2026
- AI Governance for General Counsel and Legal Teams 2026
- AI Governance for Law Firms: Privilege and Compliance 2026
- FTC AI Enforcement Actions 2026
