A State Court Just Sanctioned a Lawyer for AI Hallucinations — The Era of State-Level AI Accountability Has Arrived
Published October 9, 2025 · By The Crossing Report
Published: March 15, 2026 | By: The Crossing Report | 6 min read
Summary
A state court of appeal issued what is believed to be the first state-level appellate opinion sanctioning an attorney for AI hallucinations in March 2026. Combined with the 4th Circuit's In re Nwaubani decision the same week — a federal circuit court sanction that applied a "technology-agnostic" verification standard — the pattern is now unambiguous: AI accountability has reached the appellate level in both the federal and state court systems. For small law firm owners, this changes the risk calculus entirely. Here's why, and what to do about it.
What Happened
In the same week (March 9–15, 2026), two courts at the appellate level sanctioned attorneys for AI-related filing failures.
The 4th Circuit Court of Appeals publicly reprimanded an attorney in In re Nwaubani (2026 WL 687194) for submitting briefing with citations to nonexistent cases. The court's approach was explicitly technology-agnostic: the violation existed whether or not AI was the source, because attorneys have always had a duty to verify the accuracy of their citations. AI didn't create the obligation — it created a new way to fail it.
Days later, a California state court of appeal issued what the Daily Journal characterized as the first state appellate opinion specifically sanctioning a lawyer for AI-generated hallucinations.
These two decisions matter differently — and together, they close a gap small firm owners may have been relying on.
Why This Changes the Risk Profile
Prior to these decisions, the risk landscape looked like this:
- AI sanctions originated in federal district courts (Mata v. Avianca, SDNY 2023)
- They escalated to higher-profile federal district court decisions in 2024–2025
- Small law firm owners doing primarily state court work could reasonably argue: "I don't practice in federal court much — this is a different world"
That argument no longer holds.
State courts handle the overwhelming majority of legal work done by small law firms. Family law, estate planning, commercial disputes, employment cases, landlord-tenant, criminal defense, personal injury, contract disputes — these are state court matters. The firms that thought sanctions were a federal court problem now have a state appellate court signal to reckon with.
The combined picture from one week:
- Federal appellate courts will sanction attorneys for AI citation failures (In re Nwaubani, 4th Cir.)
- State appellate courts will do the same (California Court of Appeal, March 2026)
- The standard applied is not "did you use AI?" but "did you verify the accuracy of what you filed?"
The Escalation Pattern and What Comes Next
Sanctions patterns in court systems follow a predictable trajectory:
- Early cases → district court / trial court level warnings and sanctions
- Mid-stage cases → appellate courts apply and formalize standards
- Late stage → Supreme Court or legislative action codifies obligations
We are now in stage two simultaneously at both the federal and state levels. That means stage three — formal court rules, bar rules, or legislative requirements — is the next phase. Several bar associations are already developing guidance that formalizes verification requirements. Expect court standing orders requiring AI disclosure (already common in many federal districts) to spread aggressively to state courts in 2026.
The Greenberg Traurig eDiscovery Watch analysis published alongside these decisions documents the trajectory: courts are no longer just warning — they are sanctioning. The shift from warning to sanction is the signal.
Three Practical Moves for Your Firm
These are not future-forward recommendations. They are the minimum viable compliance posture given the current case law.
1. Verify Every AI-Generated Citation Before It Leaves the Firm
This is non-negotiable. Every case citation, statute reference, or regulatory authority generated by an AI tool must be checked against an authoritative source before inclusion in any document sent to a court, client, or opposing party.
If you use purpose-built legal AI tools (CoCounsel from Thomson Reuters, LexisNexis+ AI with Protégé, Westlaw Edge, Harvey AI), citation verification is built into the workflow. These tools are trained on legal databases and flag or ground their output in real citations.
If you use general-purpose AI tools (ChatGPT, standalone Claude, Gemini, Copilot), you are responsible for manual verification. These tools do not have access to legal databases, do not verify citations, and will confidently generate citations to cases that do not exist. The standard practice: treat every citation from a general-purpose AI as a research lead, not a filing-ready reference. Run every citation through Westlaw, LexisNexis, or Fastcase before it goes in a document.
2. Log Your Verification Process
If your verification is ever questioned — in a sanctions motion, a bar complaint, or a malpractice claim — your log is your defense.
Keep a simple record: which document, which AI tool was used, which citations were verified, by whom, and when. This does not require elaborate technology. A shared spreadsheet or a standard note in your matter management system works.
The log accomplishes two things: it disciplines your attorneys to actually run the verification step, and it creates an evidence trail showing your firm takes the obligation seriously.
3. Audit Your Court-by-Court Disclosure Obligations
More than 30 federal district courts now have AI disclosure standing orders — local rules requiring parties to disclose whether AI was used in preparing filings. State courts are adopting them at an accelerating pace.
Before your next filing in any court, check:
- Does this court have a standing order on AI disclosure?
- If yes, what does it require? (Some require disclosure only if AI was used for the substance of a brief; others require disclosure for any AI use in preparation.)
- Is your disclosure language ready?
Thomson Reuters CoCounsel and Westlaw maintain jurisdiction-specific AI disclosure trackers. If you don't have access, the court's local rules section is the starting point.
The Broader Pattern: Three Signals This Week
This week's sanctions decisions don't stand alone. Three signals from the same week point in the same direction:
Oregon HB 4154 (passed week of March 9): Added a private right of action with statutory damages for consumer-facing AI that deceives users — including client-facing tools at professional services firms.
New Hampshire SB 640 (passed committee, same week): Prohibits AI from "providing licensed professional services" without a licensed professional's meaningful oversight — with an undefined "meaningful" standard.
State court AI sanctions (Daily Journal, week of March 15): First state appellate court opinion sanctioning an attorney for AI hallucinations.
These are not random regulatory events. They are three branches of the same shift: legal accountability for AI use in professional practice is consolidating at the appellate and legislative level simultaneously.
The firms that move now — before the standard becomes a mandatory rule — are the ones that will be in a position to document compliance, not scramble to catch up.
The Bottom Line
If you are using AI tools in your practice, the compliance protocol is not complex:
- Verify citations before filing
- Log that you verified them
- Know your courts' disclosure rules
That's it. Three steps. None of them require a new technology purchase. All of them reduce your exposure to the sanction pattern now moving through both the federal and state court systems.
The attorneys who got sanctioned were not bad lawyers. They were lawyers who assumed the AI output was accurate. That assumption is now a documented liability.
The Crossing Report covers AI adoption for professional services firm owners. Subscribe for weekly intelligence on what AI means for your practice.
Related:
- Federal Appeals Courts Are Now Catching AI-Generated Filings
- Oregon Just Gave Clients the Right to Sue Over AI Chatbots
- ABA Opinion 512 AI Compliance Checklist for Law Firms
- AI Is Spawning Fake Law Firms. Here's Why That's Your Best Marketing Asset in 2026.
- AI Regulation & Compliance for Professional Services Firms — Court sanctions, bar requirements, and regulatory exposure for law firms using AI
- AI Engagement Letters & Law Firm Compliance — What engagement letters must say about AI use to satisfy ABA Opinion 512 and state bars
Frequently Asked Questions
What is the difference between the federal AI sanctions cases and the new state court sanction?
Previous AI sanctions cases — including the Mata v. Avianca case (S.D.N.Y., 2023) and the 4th Circuit's In re Nwaubani reprimand (March 2026) — came from federal courts. The March 2026 state court of appeal sanction is significant because it extends AI accountability to state court systems, which handle the vast majority of cases filed in the US each year. Most small law firm work — family law, estate planning, commercial litigation, employment, real estate — runs through state courts, not federal courts. The sanctions risk is no longer confined to firms doing federal court work.
What did the attorney do wrong in the state court AI sanctions case?
The attorney submitted legal documents containing AI-generated citations that could not be verified as real cases. This is the same core failure that produced the early federal sanctions cases: an AI language model 'hallucinated' case citations — generating plausible-sounding but nonexistent authority — and the attorney filed the document without checking whether the cases existed. In In re Nwaubani, the 4th Circuit took a 'technology-agnostic' approach: the court found a violation regardless of whether AI was explicitly identified as the source, because the attorney's duty to verify accuracy existed before AI did.
What is the minimum AI verification protocol for a small law firm?
Three steps: (1) Verify every case citation generated by AI against Westlaw, LexisNexis, or Fastcase before including it in any filing. Purpose-built legal AI tools like CoCounsel and Lexis+ Protégé have citation verification built in. General-purpose tools like ChatGPT and standalone Claude do not. (2) Maintain an internal log showing that verification was completed for any AI-assisted filing. This log becomes your evidence of due diligence if your process is ever questioned. (3) Audit your court-by-court disclosure obligations. Courts are adopting standing orders requiring AI disclosure at different rates — check the local rules for every court where you regularly file.
Which AI legal tools have built-in citation verification?
Thomson Reuters CoCounsel, LexisNexis+ AI with Protégé, and Westlaw Edge all have citation verification features that cross-check AI-generated legal analysis against their respective databases. Harvey AI performs grounding checks tied to its legal corpus. General-purpose AI tools — ChatGPT, Claude, Gemini — do not verify citations and should never be used as the sole source of case citations in any filing.
Does this affect attorneys practicing in states outside the one where the sanction occurred?
Yes. The significance of the state court of appeal sanction is not geographic — it is structural. State appellate court opinions set precedent that other state appellate and trial courts look to. The emerging pattern — from federal district courts, to federal circuit courts, now to state appellate courts — signals that every court system in the US is moving toward active enforcement. Attorneys in all 50 states should treat the verification protocol as a current obligation, not a future one.