Can lawyers use AI in New York?

What is New York’s rule on AI in court filings?

Part 161 of the Rules of the Chief Administrator of the Courts, adopted March 25, 2026 and effective June 1, 2026, is the first statewide rule, and it applies in all New York courts, civil and criminal.¹ Its baseline is permissive: using AI to prepare court papers should not be prohibited, so long as the paper meets the duties that already bind everyone who files. What the rule affirmatively requires is verification. An attorney or party who used an AI tool must carefully review the paper and independently ensure it contains no fabricated or fictitious cases, statutes, or other material. The rule is a floor rather than a ceiling: individual judges keep discretion to adopt their own AI requirements through part rules, and Part 161 supplies a model rule they can adopt. Several justices, including in Kings County and New York County, had certification requirements in their part rules before the statewide rule arrived. Check the assigned judge’s part rules every time.

Do lawyers have to disclose AI use in New York courts?

In state court, no. Part 161 expressly declined to impose a blanket disclosure mandate, reasoning that the certification duty under 22 NYCRR 130-1.1 and the candor duty under Rule 3.3 already apply to every submission however it was prepared.¹ The assigned judge’s part rules can still require it. In federal court the answer changes judge to judge: neither the Southern nor the Eastern District has a district wide rule, but individual judges do. Judge Broderick requires any filer who used generative AI to disclose the use and certify independent verification of every AI drafted portion, with a carve out for AI embedded in research platforms like Westlaw and LexisNexis.⁸ Judge Cronan publishes a model certification describing the verification steps taken. In the Eastern District, Magistrate Judge Lindsay ties an AI disclosure and accuracy certification to Rule 11(b). And Albany may yet change the state answer: Senate Bill S2698 would require an affidavit disclosing generative AI use in filings plus certification of human review; it passed the Senate Judiciary Committee in May 2025 and had not become law as of July 2026.⁹

What happens when AI fabricates citations?

New York courts sanction it, and the pace has picked up. Mata v. Avianca started the line in 2023: a $5,000 sanction, jointly and severally, plus orders to send the opinion to the client and to every judge falsely named as an author of the fake opinions.⁴ Judge Castel’s opinion opens by saying AI assistance is not inherently improper; the sanction was for the failure of the gatekeeping duty, and for standing by the fake cases after they were challenged. The Second Circuit followed in Park v. Kim, holding that Rule 11 requires an attorney to read and confirm the authorities cited, and referring the attorney to its grievance panel.⁵ The state courts have now joined. In Ader v. Ader, a Commercial Division justice awarded fees under 22 NYCRR 130-1.1 against clients and counsel jointly after a summary judgment opposition with AI fabricated citations was followed by an opposition to the sanctions motion that contained more than double the number of them.⁶ In Deutsche Bank v. LeTennier, the Third Department issued what it called the first New York appellate sanctions decision for generative AI misuse: at least twenty three fabricated cases across five filings, and $5,000 against counsel with more for the frivolous appeal.⁷ And in June 2026 the Second Department, acting on its own motion in Landberg v. City of New York, sanctioned an attorney $8,000 and his firm $2,500 for an appellate brief with fictitious Court of Appeals quotations, flagging the matter for the grievance committee.⁷ The operative rule, stated in LeTennier and adopted in Landberg: using AI never relieves the obligation to fact check and cite check every document filed.

What do the ethics rules require of a lawyer using AI?

The New York State Bar Association’s Task Force on Artificial Intelligence, whose report the House of Delegates approved on April 6, 2024, maps the duties rule by rule.² Competence under Rule 1.1 carries Comment 8’s duty to keep abreast of the benefits and risks of the technology a lawyer uses. Confidentiality is the sharpest constraint: Rule 1.6 requires informed consent before client information goes into a tool that could expose it, and the reasonable efforts duty means vetting how the vendor stores, uses, and secures the data. The Task Force advises getting assurances from the vendor even after consent. On supervision, it treats an unsupervised generative AI tool as the functional equivalent of a nonlawyer under Rule 5.3; in Mata, the nonlawyer was the tool itself. Nationally, ABA Formal Opinion 512 sets the same baseline and adds the billing rule: hourly billers charge the actual minutes spent prompting and reviewing, not the hours the task used to take.³ The opinion even anticipates where this goes: as the tools mature, lawyers may eventually have to use them to complete certain tasks competently. The guidance keeps arriving; the New York City Bar’s Formal Opinion 2025-6 now requires client notice and consent before an AI notetaker records a call. The fuller treatment of the supervision question is in Is AI allowed under the rules of professional conduct?

What does this mean for an autonomous law firm?

Every authority above converges on the same architecture: the tools may work, but the lawyer stays the gatekeeper, verifying what goes out and answering for it. That is the architecture Execute is built on. The system drafts from the case file, tracks the deadlines, and routes every document through attorney review before anything is served, filed, or sent; nothing reaches a court or a client without a lawyer’s sign off. The attorney reviews and approves. Everything else is executed. The verification the courts now demand in Part 161 and the sanctions cases is the design itself, built in by a New York trial lawyer who files under these rules himself.