AI Disclosure Rules in New York Courts

The Short Answer

New York now has a statewide rule. 22 NYCRR Part 161, adopted by Administrative Order AO/75/26 on March 25, 2026 and effective June 1, 2026, applies to every court of the Unified Court System in both civil and criminal cases. It settles two questions at once. Using an AI tool to prepare a paper is permitted, and you are not required to disclose that you used one.

What you are required to do is verify. Section 161.3 reasons that the duties owed to a court already apply to every submission whether or not AI was used, so a separate disclosure adds nothing. Section 161.4 then lets any individual court adopt a part rule of its own, and the model rule in Appendix A is explicit: an attorney who uses an AI tool must carefully review the paper and independently ensure that it contains no fabricated or fictitious cases, statutes, or other material. Your signature is the certification that you did so.

What 22 NYCRR Part 161 Actually Says

Section 161.1: It Applies Everywhere

The policy applies to all courts of the Unified Court System, in both civil and criminal cases. There is no carve-out for Supreme Court, Surrogate's Court, Family Court, or any commercial part.

Section 161.2: What Counts as AI, and What Counts as a Paper

The rule defines an AI tool broadly enough to cover the generative research and drafting products practitioners actually use. A "paper" means a brief, memorandum, affidavit, affirmation, pleading, or other document prepared for submission to a court. The definition expressly excludes materials constituting or proffered as evidence, which the rule flags as subject to separate considerations. AI-generated or AI-enhanced exhibits are a different problem with a different analysis, and Part 161 does not resolve them.

Section 161.3: No Disclosure Requirement

The operative policy statement is that AI use "should not be prohibited, as long as such use is in accordance with the duties and responsibilities that apply to individuals who submit papers to a court." Because those duties apply to every submission regardless of AI use, attorneys and parties "should not be required, upon submitting papers, to disclose to the court that they have used AI in the preparation of such papers." That language is the reason the certification forms circulating in 2024 and 2025 are no longer the statewide baseline.

Section 161.4: A Court May Still Adopt a Part Rule

Discretion remains with the individual court. Section 161.4 provides that a court may implement its own part rule on AI use and, if it does, is encouraged to adopt the model rule in Appendix A. This is the provision that matters most for a practitioner walking into an unfamiliar part, and it is covered in the next section.

Practitioner Obligations Still Come From Existing Rules

Part 161 changed the disclosure question and nothing else. The rules that produce sanctions are unchanged: New York Rule of Professional Conduct 3.3 (candor to the tribunal), 22 NYCRR 130-1.1 and 130-1.1a (frivolous conduct and the signature certification), CPLR 8303-a (costs upon frivolous claims and counterclaims), and Federal Rule of Civil Procedure 11 in federal court. Every sanctions decision in the AI hallucination cases rests on these long-standing rules.

The UCS Interim Policy on AI Is a Different Document

The Unified Court System released an Interim Policy on the Use of Artificial Intelligence on October 10, 2025. It establishes guidelines for judicial and non-judicial UCS personnel and addresses how court employees may use AI in their own work. It imposes no obligation on private attorneys and it is not the source of any filing requirement. Practitioners who encounter it should not confuse it with Part 161.

Pending Legislation Would Change the Answer

Senate Bill S2698, introduced January 22, 2025 and still in the Senate Judiciary Committee, would add a new CPLR § 2107 requiring any attorney who used generative AI to draft a document filed with a New York court to disclose that use and certify human review. As drafted, S2698 conflicts directly with the policy in Section 161.3. It has not been enacted, and until it is, the rule is Part 161.

When a Judge Adopts a Part Rule Anyway

Section 161.4 is the provision that decides what you actually have to do. Part 161 sets the statewide floor, and that floor carries no disclosure requirement. On top of that floor, any individual court may adopt a part rule of its own, and Part 161 encourages a court that does so to adopt the model rule in Appendix A rather than write its own. So compliance in a New York court is a two-step question, and the second step is the one that changes from courtroom to courtroom.

Question	Statewide, under Part 161	In a part that adopts Appendix A
May you use a generative AI tool?	Yes, under section 161.3, so long as the use is consistent with the duties that already apply to anyone submitting a paper to a court.	Yes. The model rule permits AI use and regulates what you owe the court afterward.
Must you tell the court you used one?	No. Section 161.3 states that attorneys and parties should not be required to disclose AI use upon submitting papers.	No. The Appendix A model rule does not require disclosure either. It requires review.
What must you actually do?	Verify. The duties owed to the court under the Rules of Professional Conduct and 22 NYCRR Part 130 apply to every paper whether or not AI touched it.	Carefully review the paper and independently ensure it contains no fabricated or fictitious cases, statutes, or other material.
What does your signature mean?	That the paper is not frivolous and its contentions are supported, under 22 NYCRR 130-1.1a and, in federal court, Rule 11.	That the required review was conducted. The model rule says so expressly, and failure may draw sanctions or other remedial action.
Where do you find the rule?	Part 161 applies of its own force in every court of the Unified Court System.	In the assigned judge's published part rules, which is why you pull them before you file.

Two consequences follow for an out-of-state attorney. First, there is no statewide certification form to attach and no box to check, so a filing that carries an unrequested AI disclosure is not safer than one that does not. Second, the absence of an AI provision in a judge's part rules is not permission to skip verification, because the verification duty comes from Part 130 and the Rules of Professional Conduct rather than from any AI-specific rule. The sanctions decisions below were all decided on that older ground.

The Sanctions Cases

The reported sanctions decisions share one fact pattern. An attorney used an AI tool to draft a brief, the tool fabricated case citations or quotations, the attorney did not independently verify the cited authority before filing, and the court found the hallucinations. None of them turned on a failure to disclose.

Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023)

The case that started the national conversation. Plaintiff's counsel submitted ChatGPT-generated case citations to non-existent decisions in opposition to a motion to dismiss. When the court ordered the attorneys to produce the cited opinions, they doubled down on the false authority. Judge P. Kevin Castel imposed a $5,000 sanction on the attorneys and the firm and required notice to the (fictitious) judges falsely identified as authors of the AI-generated decisions. The court found subjective bad faith based on shifting and contradictory explanations and violations of both Rule 11 and NY RPC 3.3(a)(1).

Park v. Kim, 91 F.4th 610 (2d Cir. 2024)

The Second Circuit referred counsel to the Court's Grievance Panel for further investigation after counsel filed a reply brief containing a hallucinated citation she admitted was due to AI use. The Court of Appeals stated that "citation in a brief to a non-existent case suggests conduct that falls below the basic obligations of counsel" and ordered counsel to furnish a copy of the decision to her client.

Assini v. Hayward, 2026 NY Slip Op 26086 (Sup Ct, Nassau County, June 4, 2026)

An early New York decision engaging Part 161 directly, and a useful one because it cuts in two directions. Justice Rhonda E. Fischer quashed a subpoena served on OpenAI seeking a litigant's ChatGPT prompts and outputs, holding the material protected under CPLR 3101(d) as work product. Having protected the prompts from discovery, the court then referred the litigant to 22 NYCRR Part 161 and warned that misuse of AI in preparing papers may draw sanctions. The prompts you give an AI tool are not automatically discoverable, and that confidentiality does not dilute what you certify when you sign the paper.

Practical Guidance for Out-of-State Attorneys

For attorneys appearing pro hac vice or seeking admission in a New York court, these are the practical steps that protect against sanctions and grievance referrals under Part 161 and under the rules that preceded it.

1. Pull the assigned judge's part rules before any filing. Section 161.4 leaves each court free to adopt a part rule on top of the statewide policy, so the only reliable answer is the one published by the judge who has your case. Part rules are revised regularly, and the rule in effect last month may not be the rule in effect today. State part rules are published through the Unified Court System, and each federal judge's individual practices are on the district court's website.
2. Independently verify every citation. Pull the case from Westlaw, Lexis, or PACER. Read the headnote and the cited proposition. Confirm the case stands for what your brief says it stands for. This duty exists regardless of whether AI was used.
3. Verify every quotation against the original source. Generative AI is particularly prone to fabricating block quotations that do not appear in the cited case.
4. Do not volunteer a disclosure the rules do not ask for. Section 161.3 removes any statewide duty to tell the court you used an AI tool, and the Appendix A model rule does not restore it. If the assigned judge's part rules do call for a certification, follow that text exactly rather than improvising one.
5. Document your verification steps internally. If a question is later raised, you want a clean record of when each cite was pulled, who pulled it, and what verification was performed.
6. Treat AI as draft assistance, not as a research substitute. AI tools accelerate composition. They do not authenticate sources. Every AI output is a draft to be verified, not a finding to be relied on.

Frequently Asked Questions

Does New York require attorneys to disclose use of generative AI in court filings?

No. 22 NYCRR Part 161, adopted by Administrative Order AO/75/26 on March 25, 2026 and effective June 1, 2026, applies to every court of the Unified Court System in civil and criminal cases. Section 161.3 states that attorneys and parties should not be required, upon submitting papers, to disclose to the court that they have used AI in the preparation of those papers. The reasoning is that the duties owed to a court already apply to every submission, so a separate disclosure adds nothing. What Part 161 does require is verification, and your signature on the paper is the certification that you performed it.

What is 22 NYCRR Part 161?

Part 161 is the Unified Court System's rule on the use of artificial intelligence technology, adopted by AO/75/26 on March 25, 2026 and effective June 1, 2026. Section 161.1 applies it to all courts of the Unified Court System in both civil and criminal cases. Section 161.2 defines a paper as a brief, memorandum, affidavit, affirmation, pleading, or other document prepared for submission to a court, and expressly excludes materials constituting or proffered as evidence. Section 161.3 permits AI use and imposes no disclosure requirement. Section 161.4 allows any court, in its discretion, to implement a part rule of its own and encourages the court to adopt the model rule in Appendix A.

Can a New York judge still require AI disclosure in an individual part?

Section 161.4 lets any court implement its own part rule addressing AI and encourages the court to adopt the model rule set out in Appendix A rather than write its own. The Appendix A model rule does not require disclosure either. It requires an attorney who uses an AI tool to carefully review the paper and independently ensure that it contains no fabricated or fictitious cases, statutes, or other material, and it provides that signing the paper certifies that the review was conducted. Because a court may adopt something other than the model rule, the reliable answer is always the assigned judge's published part rules, pulled before the filing.

What happens if AI fabricates a citation in a New York filing?

Sanctions, struck filings, and referral to grievance authorities. In Mata v. Avianca (S.D.N.Y. 2023), Judge P. Kevin Castel sanctioned the attorneys and their firm $5,000 for submitting fabricated AI-generated case citations and then standing behind them. In Park v. Kim (2d Cir. 2024), counsel was referred to the Court's Grievance Panel after filing a reply brief containing a hallucinated citation. These cases were decided on Rule 11, 22 NYCRR Part 130, and Rule 3.3 of the Rules of Professional Conduct rather than on any AI-specific rule, which is why the verification duty matters more than the disclosure question.

What do I need to do before filing AI-assisted papers in New York?

First, pull the assigned judge's part rules, because section 161.4 leaves each court free to adopt its own. Second, independently verify every citation and every quotation against the original source. Fabricated authority is the single largest source of sanctions in this area. Third, do not volunteer a disclosure that no rule asks for, and if a part rule does call for a certification, follow that text exactly. Fourth, keep an internal record of the verification steps you took, including when each cite was pulled and by whom, in case the question is raised later.

Does the New York Unified Court System Interim Policy on AI apply to private attorneys?

No. The UCS Interim Policy on Artificial Intelligence, issued October 10, 2025, governs the responsible use of AI by judicial and non-judicial UCS personnel. It is an internal policy and imposes no obligation on private attorneys filing papers in New York courts. It is a different document from 22 NYCRR Part 161, which is the rule that reaches attorneys and parties. Practitioner obligations come from Part 161, any part rule adopted under section 161.4, the New York Rules of Professional Conduct, 22 NYCRR Part 130 on frivolous conduct, and Rule 11 in federal court.

Are my AI prompts discoverable in New York litigation?

Not automatically. In Assini v. Hayward, 2026 NY Slip Op 26086 (Sup Ct, Nassau County, June 4, 2026), the court quashed a subpoena served on OpenAI seeking a litigant's ChatGPT prompts and outputs, holding the material protected as work product under CPLR 3101(d). The same decision then referred the litigant to 22 NYCRR Part 161 and warned that misuse of AI in preparing papers may draw sanctions. The prompts are shielded from discovery, and that shielding does not reduce what the signer certifies about the finished paper.

Related: Pro Hac Vice Overview · How to File a PHV Motion · PHV Application Guide · PHV Cost · PHV by Order to Show Cause · PHV in SDNY and EDNY · PHV in the Commercial Division

This article is general information, not legal advice. It describes 22 NYCRR Part 161 as adopted by AO/75/26 and effective June 1, 2026. Individual part rules adopted under section 161.4 vary from court to court and are revised frequently, so confirm the current rules of the assigned judge before any filing. Attorney advertising.