New York state court officials have released a new guide telling attorneys they can use generative artificial intelligence tools to help prepare court papers, and they don't have to tell a judge they did it. The catch: lawyers are fully on the hook if that AI-generated material turns out to be wrong, fabricated, or misleading.
The guide, titled The Risks of Using Artificial Intelligence in Preparing Court Papers: What Every Attorney Should Know, was developed by the court system's Advisory Committee on Artificial Intelligence and the Courts, according to nydailyrecord.com. It spells out that attorneys who use generative AI tools must understand the capabilities and limitations of those tools and independently ensure the accuracy and reliability of any information the software spits out. As amNewYork reports, attorneys are not required to disclose AI use at the time of filing or provide separate certifications beyond what they already sign off on.
That no-disclosure approach has at least one legal educator uneasy. Heidi Brown, an associate dean at New York Law School, said she's concerned the guidance does not require attorneys to disclose AI use to judges, and that disclosure and certification requirements could actually encourage attorneys to check their AI-assisted work more carefully, per amNewYork's reporting.
What the Guidance Actually Requires
Despite skipping the disclosure mandate, the guidance is blunt about where responsibility lands. Attorneys remain fully responsible for ensuring that every factual and legal assertion submitted to a court is accurate and properly supported, the document states, as reported by nydailyrecord.com. It urges lawyers to confirm all AI-generated citations and quotations through authoritative legal research tools and to document the steps they took to verify material produced with AI assistance.
Per amNewYork, the guidance also makes clear that attorneys remain subject to the same responsibility and sanctions for false material in court filings whether or not AI was used, and that using AI does not alter an attorney's obligations under the Rules of Professional Conduct or 22 NYCRR Part 130, according to nydailyrecord.com. The guidance advises attorneys to independently search for, read and analyze any AI-generated citations rather than taking the software's output at face value.
Part of the concern stems from how these tools actually work. The guidance explains that most publicly available large language models, including ChatGPT, Claude, Gemini and Grok, do not consult databases of verified facts or confirm that citations correspond to real cases, per amNewYork. AI models generate plausible-sounding text by predicting likely next words rather than retrieving verified information, which is precisely why independent verification matters.
'Misgroundings' Seen as a Subtler Danger
Brown told amNewYork that so-called misgroundings — AI citations of real cases or statutes that inaccurately describe their meaning, omit nuance, misstate relationships among cases, or lack current information — can be more dangerous than straightforward AI hallucinations, since they look legitimate on the surface. She said attorneys must verify that every sentence and proposition is actually supported by the citations AI provides, warning that checking AI research thoroughly can take substantial time and that newer attorneys may have difficulty identifying errors. Brown also noted that newer attorneys may not even realize that the legal research platforms they already use have AI embedded in them, or that their work already contains AI-generated inaccuracies.
The guidance also flags bias risks baked into the technology. Training datasets may contain cultural, economic and social biases and prejudice against protected classes, producing harmful or biased content, the document warns, and AI can draw only on its training data, meaning it may not reflect later developments in the law.
How the New Guidance Fits Into Part 161
This latest guide builds on a formal rule the court system adopted earlier this year. Part 161 of the Rules of the Chief Administrator of the Courts was adopted March 25, 2026 and took effect statewide June 1, 2026, becoming New York's first statewide AI rule applying in all courts, civil and criminal, according to execute.law. Under that rule, attorneys may use AI in preparing court papers with no blanket disclosure required, but the attorney who used it must independently verify that nothing filed contains fabricated or fictitious cases, statutes, or other material, per the same source.
Individual judges still retain discretion to adopt their own AI requirements through their part rules, execute.law notes, meaning some courts could impose disclosure obligations the statewide guidance does not. New York State Chief Administrative Judge Joseph Zayas, who appointed the Artificial Intelligence Advisory Committee in 2024, said AI can improve efficiency but requires caution and informed oversight, according to amNewYork's reporting. The court system decided not to ban generative AI outright or mandate disclosure, instead betting on informed and responsible use by the attorneys themselves.
Sanctions Are Already Piling Up Nationally
The stakes behind all this guidance are not theoretical. Court sanctions against attorneys for filing briefs containing AI-generated errors took off last year and the rate is still increasing, according to NPR. A researcher tracking the problem has counted more than 1,200 cases involving erroneous AI-generated information to date, roughly 800 of them from U.S. courts, per NPR's reporting.
New York has already seen its own fallout. In January 2026, the Appellate Division, Third Department, imposed monetary sanctions on counsel and a litigant over an opening brief and successive filings that together cited 23 fabricated cases, according to the New York State Bar Association. That case followed a now-familiar pattern set in June 2023, when Judge P. Kevin Castel imposed a $5,000 sanction on lawyers and their firm in Mata v. Avianca after fabricated, ChatGPT-generated cases turned up in a filing, the Bar Association notes. In January 2024, the Second Circuit went further, referring attorney Jae S. Lee to its grievance panel for citing a nonexistent AI-generated case.
Penalties elsewhere have grown steeper still. An Oregon federal court may have set a record with an order for a lawyer to pay $109,700 in sanctions and costs for AI-generated errors, NPR reports, while lawyers for MyPillow CEO Mike Lindell were fined $3,000 each over briefs containing fictitious AI-generated citations. In March 2026, the 6th U.S. Circuit Court of Appeals sanctioned attorneys Van Irion and Russ Egli in Whiting v. City of Athens after finding more than two dozen fake citations and factual misrepresentations in an appeal showing signs of AI hallucinations, according to Reuters; the pair were ordered to reimburse the opposing city for its legal work and each pay the appeals court $15,000 as a punitive sanction.
With AI tools now being built directly into much of the software lawyers already use, as NPR has reported, New York's approach puts the burden squarely on individual attorneys to catch what the machines get wrong — before a judge, or an opposing counsel, catches it for them.
