New York’s Unified Court System has issued a seven-page guide for lawyers preparing court papers with generative AI, Bennett Loudon reported on 27 October. Developed by the Advisory Committee on Artificial Intelligence and the Courts, the guide permits AI-assisted drafting without a disclosure or separate certification at filing. Lawyers must independently check the material they submit, under a model rule the court system adopted earlier in 2026, and remain responsible for their papers under existing rules.
Key points
- New York’s guidance permits generative AI in court papers without requiring lawyers to disclose its use or file an extra certification.
- An earlier model rule requires lawyers using AI to understand its limitations and independently verify the information it produces.
- The guide urges checks of citations, quotations and bias. Existing professional conduct rules and 22 NYCRR Part 130 continue to apply.
- New York Law School Associate Dean Heidi Brown argues that disclosure could prompt lawyers to check AI-assisted work more carefully.
The model rule and 22 NYCRR Part 130
The guide is titled The Risks of Using Artificial Intelligence in Preparing Court Papers: What Every Attorney Should Know. Chief Administrative Judge Joseph A. Zayas appointed the advisory committee to examine AI’s role in the judiciary, according to Loudon. The court system’s judicial leadership approved the committee’s document for public distribution. It is guidance for attorneys using the technology, alongside a model rule the court system had already adopted.
That model rule requires attorneys who use generative AI to understand what the tools can and cannot do, and to independently establish the accuracy and reliability of the information they generate. The new guide describes how to apply that obligation when preparing filings. It does not add an AI-use declaration to the filing process: signing a paper remains the lawyer’s certification, Loudon reported.
The underlying duties extend beyond AI output. The Rules of Professional Conduct still govern attorneys’ submissions, while 22 NYCRR Part 130 governs sanctions for frivolous conduct in New York civil litigation. The guide says lawyers remain responsible for accurate, properly supported factual and legal assertions in papers submitted to a court. Those rules, rather than a new AI-specific disclosure form, are the route by which the court system holds an attorney to account for a deficient filing.
Checking citations in New York court papers
The guide urges lawyers to check AI-generated citations and quotations against authoritative legal research sources, examine generated text for bias or inappropriate material, and document their verification steps. The checks apply to the content of a filing even though a lawyer need not tell the judge whether AI helped draft it.
Courts around the country have encountered filings with invented case citations and inaccurate quotations produced by AI systems. Joseph Zayas, the chief administrative judge, said the tools could improve efficiency but carried risks requiring oversight. “AI tools can help enhance efficiency, but they also carry risks that demand caution and informed oversight,” he said, according to Finger Lakes Daily News. The court system’s approach leaves attorneys able to use those tools while making independent review a condition of doing so.
The review must go further than finding out whether a cited case exists. The guidance calls for lawyers to establish that citations are accurate and relevant by finding, reading and analysing the underlying legal material themselves, amNY reported. It also describes how a model can produce plausible text without checking it against verified facts. A real citation therefore still requires a lawyer to examine the proposition for which it is being used.
Heidi Brown argues for disclosure
Heidi Brown, a New York Law School associate dean, called the absence of a disclosure requirement “concerning”. She argued that requiring lawyers to declare AI use and make an additional certification might make them more attentive to checking their work, amNY reported. Brown also raised a difficulty with legal research platforms that incorporate AI: a newer lawyer might fail to recognise that material included in a filing was generated by a model.
Brown’s concern includes what she calls “misgroundings”, in which a tool supplies a genuine case or statute but describes it inaccurately, misses a qualification or overlooks a later development. She said lawyers need to check the propositions attached to citations, as well as the citations and quotations themselves. That work, Brown told amNY, is time-consuming, and she said newer attorneys can find these errors harder to catch.
