A Court Just Issued the Largest AI Sanctions in U.S. History — $110,200. Here's the Verification Protocol Every Law Firm Needs Before the Next Filing

April 21, 20266 min readBy The Crossing Report

A Court Just Issued the Largest AI Sanctions in U.S. History — $110,200. Here's the Verification Protocol Every Law Firm Needs Before the Next Filing

On April 4, 2026, U.S. Magistrate Judge Mark Clarke sanctioned two San Diego attorneys a combined $110,200 for submitting AI-fabricated case citations to a federal court. The lead attorney, Stephen Brigandi, was ordered to pay $96,000. Co-counsel Tim Murphy was ordered to pay $14,200.

The filings contained 23 fabricated case citations and 8 false quotations from cases that do not exist. The AI generated them. The attorneys filed them.

Their client's case — a family business dispute over Valley View Winery — was dismissed with prejudice. It cannot be refiled.

This is now the largest AI sanctions award in U.S. history. It will not be the last.


What Actually Happened

Between January and March 2025, Brigandi filed three motions in a family business case. All three contained citations to cases that do not exist — plausible-looking case names, realistic citation formats, quotes from rulings that were never issued. The citations were generated by AI. Neither attorney verified them before filing.

The court identified the fabrications. By April 2026, the sanctions order landed: $110,200 total, split between the two attorneys. The family lost their entire legal case as a direct result.

The scale of this single case is worth pausing on. HEC Paris researcher Damien Charlotin, who maintains what is likely the most comprehensive database of AI hallucination cases in legal proceedings, now tracks 1,200+ such cases globally. Approximately 800 are from U.S. courts.

The sanction trajectory:

  • 2023: Warnings, verbal reprimands, fines under $1,000
  • 2024: Fines of $1,000–$5,000 per case
  • 2025: Fines reaching $10,000+
  • Q1 2026 alone: $145,000+ in total attorney sanctions across multiple cases

The Brigandi case accounts for $110,200 of that Q1 2026 total. The escalation is not slowing.


The Judicial Double Standard — And Why It Doesn't Change Your Risk

Here's the piece of context that creates the most heat in legal circles: while courts are sanctioning attorneys for AI use at escalating rates, judges are using AI too — with minimal oversight.

A 2026 investigation found that 61.6% of federal judges use AI tools in their judicial work. Of those judges, 45.5% received no formal training from their courts on AI use.

The asymmetry is real. Courts are not opposed to AI adoption. They are running a two-tiered enforcement posture: aggressive sanctions for attorneys who use AI in ways that produce fabricated filings, minimal accountability for judicial AI use. The argument circulating in legal reform circles is that this isn't primarily about ethics — it's about professional gatekeeping and institutional authority.

That argument may be correct. But it does not reduce your firm's exposure by one dollar.

The enforcement campaign is ongoing. The sanctions are real. The clients at risk are yours. The right response to an unfair enforcement asymmetry is not to skip the verification step — it is to have the verification protocol that protects your clients and your practice while the systemic issue works itself out in the courts and bar associations.


The Three-Step Verification Protocol

This is the minimum standard for any law firm using AI in client work that will be submitted to any court.

Step 1 — Verify every citation, every time

Every case cited in any court document must be verified against Westlaw, LexisNexis, Fastcase, or another authoritative legal database before submission. No exceptions for "quick filings" or "obvious cases."

The tool you use for drafting matters. General-purpose AI tools — ChatGPT, Claude standalone, Google Gemini — do not verify citations against real case databases. They generate text that looks like real citations. Purpose-built legal AI tools with integrated citation verification (Thomson Reuters CoCounsel, LexisNexis+ AI with Protégé, Harvey AI) perform grounding checks tied to their legal corpora. If your firm uses general-purpose AI for any step in document drafting, the verification step must happen before the document leaves the firm.

Step 2 — Log the verification

For every filing where AI-assisted drafting was used, maintain an internal record showing that citation verification was completed. A simple log — matter number, filing date, citations checked, verified by — is sufficient. This record becomes your evidence of due diligence if your process is ever challenged.

This is not about compliance theater. If a court questions your process, the log demonstrates that the verification step was real and systematic — not something you claimed retroactively after a problem was identified.

Step 3 — Name the person responsible

One named person on every matter is responsible for citation verification before any AI-assisted filing is submitted. This cannot be delegated back to an AI tool, left as a shared responsibility, or handled through a general reminder at the end of the review process.

The assignment is explicit: this person checks every citation, confirms they exist in an authoritative database, and signs off before submission. For a solo practice, that person is the attorney of record. For a small firm with associates or paralegals, the assignment must be made and documented.


The Part of This Story That Gets Underreported

The $110,200 in sanctions is the number that gets the headline. The dismissal with prejudice is the consequence that matters more.

Brigandi and Murphy's client lost their legal case entirely. The dismissal with prejudice means the family business dispute cannot be refiled. Whatever claim they had — whatever remedy they were pursuing — is extinguished. Not delayed. Not restarted with a different attorney. Gone.

The attorney's error destroyed a client's legal rights.

For a law firm owner, this is the risk calculation that should govern every decision about AI use in client work. The financial exposure from sanctions is real. The exposure from malpractice claims when a client's case is lost due to AI-generated errors is potentially larger. The reputational and relationship damage — a client whose legal rights were destroyed because you didn't verify a citation — is the exposure that doesn't appear in sanctions databases.

The verification protocol is not about compliance. It is about protecting your clients from outcomes you cannot undo.


One Step to Take This Week

If your firm uses AI in any drafting that produces court filings, do two things before your next submission:

1. Confirm your verification workflow is explicit, not assumed. Is there a named person responsible for verifying citations on every filing? Is it written down? If your current practice is "we check citations before submitting" without a specific person assigned and a log maintained — you have a gap.

2. Check your AI tool's citation verification capability. If you use a general-purpose AI tool (ChatGPT, Claude, Gemini) for any part of legal drafting, note that citation verification requires a separate step against an authoritative legal database. Purpose-built legal AI tools (CoCounsel, LexisNexis+, Harvey) include this verification. Know which tools in your stack verify citations and which do not.

The Brigandi case won't be the last record-setting sanctions order. But it can be the last one that catches your firm without a protocol.

Frequently Asked Questions

What happened in the Brigandi AI sanctions case?

On April 4, 2026, U.S. Magistrate Judge Mark Clarke sanctioned San Diego attorney Stephen Brigandi $96,000 and co-counsel Tim Murphy $14,200 — totaling $110,200 — for submitting three motions containing 23 fabricated case citations and 8 false quotations generated entirely by AI. The citations were fake cases with plausible names, realistic citation formats, and invented quotes from fictional rulings. As a direct result, their client's entire case — a family business dispute over Valley View Winery — was dismissed with prejudice, meaning it cannot be refiled. It is the largest AI sanctions award in U.S. history as of April 2026.

How common are AI hallucination sanctions cases in U.S. courts?

More common than most attorneys realize. HEC Paris researcher Damien Charlotin maintains the world's most comprehensive database of AI hallucination cases: as of April 2026, there are 1,200+ documented cases globally, approximately 800 from U.S. courts. The pattern shows clear escalation: early cases (2023) resulted in warnings or small fines under $1,000. By 2024, sanctions reached $1,000–$5,000. By 2025, $10,000+. In Q1 2026, $145,000+ in total attorney sanctions across multiple cases. The Brigandi case alone is $110,200 — more than the entire Q1 2025 total.

Does the judicial double standard matter for my law firm's risk calculation?

It matters as context, but it does not reduce your risk. A 2026 investigation found that 61.6% of federal judges use AI tools in judicial work, with 45.5% of those judges receiving no formal training from their courts. Courts are not opposed to AI — they are enforcing accountability asymmetrically: attorneys are being sanctioned at escalating rates while judicial use of AI is minimally supervised. For a small firm, this asymmetry has one practical implication: the enforcement campaign is ongoing, the sanctions are real, and the cost-benefit analysis favors having a verification protocol rather than debating the fairness of the enforcement environment.

What is the three-step AI citation verification protocol for law firms?

Step 1 — Verify every citation before filing: Every case cited in any document that will be submitted to a court must be verified against Westlaw, LexisNexis, Fastcase, or another authoritative legal database. General-purpose AI tools (ChatGPT, Claude standalone, Gemini) do not verify citations. Purpose-built legal AI tools (Thomson Reuters CoCounsel, LexisNexis+ AI with Protégé, Harvey AI) include citation verification in their workflow. Step 2 — Log the verification: Maintain an internal record showing that verification was completed for each filing where AI-assisted drafting was used. This becomes your evidence of due diligence if your process is ever challenged. Step 3 — Assign explicit responsibility: One named person on every matter is responsible for citation verification before submission. This cannot be delegated to an AI tool or left as a general expectation. Name the person, document the check.

Is the verification obligation the same for small law firms as for BigLaw?

Yes. The 'non-delegable duty to verify' standard applies regardless of firm size. A solo practitioner using AI has the same verification obligation as a 500-attorney firm. The Brigandi case involved two attorneys at what appears to be a relatively small practice — not a major firm with a compliance department. Small firms have the same exposure with none of the compliance infrastructure that larger firms have developed. The practical consequence: small firms need a written, practiced verification protocol at least as much as large firms do.

What happens to the client when a lawyer's AI-generated citations are sanctioned?

In the Brigandi case, the client's entire case was dismissed with prejudice. This is the outcome that rarely gets sufficient attention in the coverage of attorney sanctions: the lawyer faces a financial penalty, but the client loses their legal matter entirely. A dismissal with prejudice means the case cannot be refiled — the client's legal claim is extinguished. For a law firm owner, this outcome is the most important part of the Brigandi story: the attorney's error wasn't just a professional embarrassment or a financial penalty. It destroyed a client's legal rights.

Get the weekly briefing

AI adoption intelligence for accounting, law, and consulting firms. Free to start.

Related Reading

This is the kind of intelligence premium subscribers get every week.

Deep analysis, cross-sector patterns, and the frameworks that help professional services firms make the crossing.