New York’s New AI Court Rule Leaves Both Sides of a Case Exposed

New York’s court system began enforcing Part 161 on June 1, permitting attorneys to draft court filings with artificial intelligence without ever telling a judge they did it. Nothing in the rule bans the technology or forces anyone to admit using it.

Twelve days earlier, a Second Department appellate judge had already shown who else can be blamed when an AI-written citation turns out to be fake: the lawyer across the table who never touched the software.

The Mechanics of Part 161

Part 161 covers what the rule defines as a paper, meaning pleadings, motion briefs, memoranda, affidavits and similar documents filed or served in a case, though not anything offered as evidence.

It defines artificial intelligence broadly, as any machine-based system that makes predictions, recommendations or decisions from human-defined objectives and inputs. Attorneys and parties can use that kind of tool to prepare those documents, as long as the work still meets the duties that applied before AI existed.

Nobody has to admit it. The rule does not require attorneys to disclose AI use to the court at all, under its own text.

That freedom comes with a string attached. Attorneys are required to independently review and verify the accuracy of anything AI helped write before a judge sees it. The obligation is not new. It runs through 22 NYCRR 130-1.1a, which turns an attorney’s signature into a certification that a filing is not frivolous, inaccurate or misleading, and through Rule 3.3 of the state’s Rules of Professional Conduct, which bars a false material factual statement or a frivolous legal argument.

An Appellate Judge Puts Both Sides on Notice

Twelve days before Part 161 took effect, a Second Department panel handed down a preview of exactly the problem the rule was built around.

In Landburg v. City of New York, the appellant’s attorney had filed papers loaded with fake case citations. The attorney on the other side never flagged them either.

If you read the brief and looked at the cases, you would have realized it was your responsibility

Justice Valerie Brathwaite Nelson of the Appellate Division’s Second Department said that to the respondent’s lawyer during the May 20 argument, according to New York attorney and legal journalist Nicole Black, writing for the Daily Record.

It was not an isolated warning. A Seventh Circuit panel made a similar point in March, in a case called Dec v. Mullin, criticizing an attorney for failing to catch his own opponent’s hallucinated cases.

Part 161’s certification duty technically binds only the attorney who signs the paper. New York courts are already treating the lawyer who reads that paper from across the table as accountable too.

Sanctions Pile Up Nationwide

The pattern started in 2023, when a lawyer’s ChatGPT-drafted brief in Mata v. Avianca cited six cases that never existed, drawing a $5,000 sanction that looked shocking at the time. It no longer does.

US courts imposed more than $145,000 in sanctions for fabricated AI citations in the first quarter of 2026 alone, nationwide, according to an accounting of the quarter’s rulings.

Fortune reported the Oregon case as the largest AI hallucination penalty in American legal history to date, a record that lasted only months before the next filing season.

Legal researcher Damien Charlotin’s tracker has logged more than 1,700 rulings worldwide as courts keep finding new cases. More than 1,000 of those were in the United States as of May 2026, the same tracker showed.

The newest addition came from the Ninth Circuit. In LNU v. Blanche, the panel’s June 3 sanctions order punished two attorneys for briefs full of nonexistent cases, misattributed quotations and what the judges called gross misrepresentations of real cases. The panel was careful about where the misconduct actually occurred: sanctions attach at the point of signing and filing a brief, not simply for using generative AI to research or draft one.

One reason the errors keep surfacing: a study cited in a related sanctions decision found legal research platforms produced hallucinations in 17% to 33% of their responses. AI meeting notetakers elsewhere have already turned from time-saver into legal liability, the same underlying pattern now showing up in litigation.

A Statewide Floor With a Judge-by-Judge Ceiling

Part 161 sets a floor, not a ceiling. Individual judges can adopt the model rule in Appendix A, write their own AI order, or do nothing at all.

A court that adopts Appendix A turns a signature into an explicit certification that a filing contains no fabricated cases, statutes or other material. That certification is per-court, not statewide. Many individual New York state and federal judges have already gone further than Part 161 on their own, and requirements vary significantly by judge.

Attorneys are left with homework the rule does not do for them:

• Verify independently – confirm every AI-suggested citation and quote against the primary source before it reaches a filing.
• Check the judge’s part rules – confirm at case intake whether the assigned court adopted Appendix A or issued its own AI order.
• Track Senate Bill S2698 – watch whether Albany moves from implicit certification to mandatory disclosure.

Florida picked the opposite design. Its Supreme Court order bars local circuits from requiring AI disclosure, trading New York’s patchwork for one uniform statewide standard effective June 15.

India’s Supreme Court chose a stricter posture still, moving to bar algorithms from the bench entirely, a sharp contrast with New York’s opt-in, attorney-facing approach.

Are the Judges Using AI Too?

Yes. A Northwestern University survey of federal judges, published by the Sedona Conference in March 2026, found that most of the bench already uses AI tools in some capacity, even as those same judges sanction lawyers for the technology’s mistakes.

• 60%+ of responding federal judges said they use AI tools in some capacity.
• About 25% of judges formally permit AI use in chambers, while roughly 20% ban it outright.
• Nearly half of respondents said their court administration gave them no AI training at all.

A survey of 502 federal judges, conducted by Northwestern University, produced those figures, and the American Bar Association flagged the same asymmetry: courts enforcing verification standards on lawyers that much of the bench has not fully applied to itself.

New York’s own judges operate under a separate track. The Unified Court System issued an Interim Policy in October 2025 limiting judicial and staff AI use to pre-approved programs and treating AI strictly as an assistant, never a substitute for a judge’s own decision. Employment courts are absorbing a fresh wave of AI workplace lawsuits built on a similar tension between efficiency and verification.

A Disclosure Bill Waits in Albany

New York’s legislature already has a bill that would flip Part 161’s entire approach.

Senate Bill S2698 would require any attorney who used generative AI to draft a court filing to disclose that use and certify that a person reviewed and verified the content. It would add a new CPLR section requiring a disclosure affidavit in civil filings, plus a matching requirement for appellate briefs.

It remains parked in the Senate Judiciary Committee, with no floor vote scheduled. If it passes, New York moves from a signature that implies verification to a form that spells it out.

Frequently Asked Questions

Does Part 161 require New York attorneys to disclose AI use in court filings?

No. Part 161 sets no statewide disclosure duty, though individual judges can demand it through their own part rules. Nationally, at least 25 federal district courts have already adopted standing orders requiring attorneys to certify whether AI was used in a filing, a stricter posture than New York’s floor.

What happens if a New York filing contains a fabricated AI citation?

The signing attorney can face sanctions, fee-shifting or a referral to a grievance committee for discipline, on top of the underlying argument getting stricken. One New York court referred an attorney to the local Attorney Grievance Committee after finding Rules of Professional Conduct violations tied to unreviewed AI content.

Does Part 161 cover evidence like bodycam footage or recordings?

No. The rule applies only to papers such as briefs, motions and affidavits. Audio recordings, depositions, bodycam footage, jail calls and witness interviews fall under separate discovery and evidentiary rules instead.

Can a New York judge require more than Part 161 demands?

Yes. New York’s Commercial Division Advisory Council has recommended its own Rule 6(e), which would make anyone who files material certify its accuracy and reliability, on top of whatever Part 161 already requires.

Are New York’s own judges allowed to use AI?

Yes, under a separate Interim Policy the Unified Court System issued in October 2025 for judges and non-judicial staff. That policy limits use to pre-approved programs, including Microsoft Copilot and OpenAI’s ChatGPT, and treats AI strictly as an assistant.

What would Senate Bill S2698 change if it passes?

It would require an attorney who used generative AI to draft a civil filing to submit an affidavit disclosing that use and certifying that a person reviewed and verified the content, plus a matching disclosure requirement on appellate briefs.

Disclaimer: This article is for informational purposes only and does not constitute legal advice. Court rules, standing orders and sanctions practices vary by judge and jurisdiction and can change over time. Attorneys should consult the official rule text or qualified counsel for guidance on a specific matter. Figures are accurate as of publication.