New York courts issue generative AI guidance without disclosure mandates

New York state court administrators have issued guidance warning legal practitioners against the unverified use of generative artificial intelligence in court filings.

The advisory addresses mounting judicial concern across the legal system regarding algorithmic tools fabricating case law, citations, and statutory arguments. Attorneys and self-represented litigants must independently verify the accuracy of every authority, factual assertion, and quotation before lodging submissions.

The guidance highlights the structural hazards of generative models, particularly their tendency to produce hallucinated citations that appear authentic. Language models generate plausible-sounding text based on statistical probability rather than verified legal databases, often inventing docket numbers, judicial opinions, and appellate precedents. Court authorities emphasised that human review must validate every reference against authoritative legal reporters or official databases.

Despite these operational warnings, the guidance stops short of imposing a blanket requirement to disclose generative AI use in court submissions. State judicial authorities declined to mandate formal declarations detailing whether practitioners employed algorithmic assistants during drafting. This approach contrasts with several federal jurisdictions where judges have issued standing orders requiring explicit disclosure whenever software contributes to a brief.

Court administrators concluded that existing professional conduct standards and procedural rules are sufficient to govern generative systems. Under established New York disciplinary codes, attorneys remain strictly accountable for all documents submitted under their signature. Submitting non-existent authorities remains subject to sanctions, irrespective of whether an error originated from a colleague or an automated model.

The advisory recognises that generative tools offer productivity gains in legal research, drafting, and case synthesis. However, it makes clear that practitioners cannot delegate professional competence or supervisory duties to synthetic systems.

By relying on established conduct rules rather than bespoke disclosure forms, New York relies on practitioner compliance to preserve filing integrity. Will traditional disciplinary oversight prove adequate as generative writing tools integrate deeper into commercial legal workflows?