AI Regulation Tracker / Litigation
AI Sanctions Reach the State Courts: A New York Appellate Panel Fines a Firm for Fake Citations
New York's Appellate Division, Second Department, has started imposing financial sanctions for AI hallucinated citations, including a recent $2,500 penalty against a personal injury firm. The federal warnings have arrived in state appellate practice. Here is what every filing now has to survive.
The warnings reached the state courts
For most of the past year the cautionary tales about AI in legal filings came from federal courtrooms. A published Ninth Circuit opinion. A Mississippi judge who removed every lawyer from a case. A federal appeals panel that sanctioned two lawyers for briefs full of cases that did not exist. Those stories let a lot of state-court practitioners file the problem under "someone else's jurisdiction." That filing is now closed.
New York's Appellate Division, Second Department, the busy intermediate appellate court based in Brooklyn, has begun handing out financial sanctions for AI hallucinated citations. In a June 2026 order the court directed the Law Offices of Michael Lamonsoff to pay a $2,500 sanction, and it imposed a separate $8,000 penalty in a related matter, as reported by the New York Law Journal on June 24, 2026. Weeks earlier, in May 2026, the same court sanctioned a self-represented filer over an AI error and held that pro se status was no excuse.
Why a small fine is a big signal
A $2,500 sanction will not bankrupt a law firm. That is exactly why it matters. The number is not the point. The precedent is. When an intermediate appellate court, the level where most litigators actually practice, starts treating a fabricated citation as sanctionable conduct rather than an embarrassing mistake, the calculus for every filer in the state changes.
Three signals sit inside these orders. The first is that the duty to verify belongs to the person who signs the brief, not to the tool that drafted it. The second is that the firm name carries the exposure, so a junior associate's unverified draft becomes the firm's problem. The third, from the pro se ruling, is that the standard does not bend for the unrepresented. The court is describing a floor that applies to everyone who files.
The pattern behind the New York orders
New York is not an outlier. It is a data point in a curve that is bending upward. The AI Hallucination Cases Database maintained by Damien Charlotin, which tracks court decisions addressing AI fabrications in filings, has logged a steady rise in sanctions across federal and state courts through 2026. The New York orders fit a national pattern in which judges have moved from warning to penalty, and from federal trial courts to state appellate benches. A litigator watching only the federal headlines has been watching the wrong half of the map.
What the orders ask of a filing
None of these rulings break new doctrinal ground. They apply an old duty to a new failure mode. A lawyer has always been responsible for the accuracy of the authorities cited to a court. Generative AI did not change that duty. It created a fast, confident way to violate it, by producing citations that look real, read real, and do not exist. The court's response is to enforce the duty that was always there.
The practical reading is narrow and strict. Every authority in a brief must be a real decision that says what the brief claims it says. An AI-assisted draft is not finished when it reads well. It is finished when a human has pulled each cited case from a primary database and confirmed it. The gap between those two states is exactly where the sanctions live.
The workflow that survives
The defense is a verification habit, documented. Adopt one firm-wide rule: no citation leaves the office until a person has retrieved and read the actual decision. Keep a short record that the check happened, because if a citation is ever challenged, the existence of a verification step is part of the answer. Treat AI output as a research lead, never as a citation of record. And brief everyone who touches a filing, including contract attorneys and self-represented clients a firm assists, that the unverified citation is the liability, not the AI.
The firms that build this habit will keep the speed that AI offers and shed the risk. The ones that keep pasting confident citations into briefs are now doing it in a state where the appellate court has shown it will write a check in their name.
Frequently Asked Questions
Did a New York court really sanction a firm for AI citations?
Yes. New York's Appellate Division, Second Department directed the Law Offices of Michael Lamonsoff to pay a $2,500 sanction in a June 2026 order, with a separate $8,000 sanction in a related matter, as reported by the New York Law Journal on June 24, 2026. The same court imposed a sanction on a self-represented filer over an AI error in May 2026.
Is this different from the federal AI sanctions cases?
It extends the same pattern into state appellate practice. Federal courts, including the Ninth Circuit and a Mississippi federal judge, had already sanctioned lawyers for AI hallucinations. The New York orders show intermediate state appellate courts now doing the same, which reaches the level where most litigators practice.
Who is responsible when AI invents a citation?
The filer. The orders treat the duty to verify cited authority as belonging to the attorney or litigant who signs the brief, not the tool. The firm whose name appears on the filing carries the exposure, and the court held that self-represented status does not lower the standard.
How do I keep using AI without this risk?
Use AI as a research lead, never as a citation of record. Require that a human retrieve and read every cited decision from a primary database before filing, and log that verification. The fabricated citation, not the use of AI, is what draws the sanction.
Sponsored Training
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.