New York Courts Mandate Verification for AI Filings

New York court administrators have issued formal guidance governing the use of generative artificial intelligence in legal submissions across the state.

The advisory targets the growing integration of large language models into legal drafting and research. It warns practitioners of systemic risks, focusing on fabricated authorities, factual inaccuracies, and breaches of client confidentiality.

Under the new framework, lawyers must independently verify every citation, quotation, and legal argument derived from generative AI tools against authoritative sources. Court administrators reiterated that existing rules of professional conduct apply to AI-assisted work. The physical or electronic signature on a court paper constitutes a personal certification of its accuracy. Counsel cannot shift accountability to software providers when filings contain fictitious precedents or mischaracterised statutory provisions.

The guidance deliberately stops short of imposing a blanket disclosure requirement. Attorneys are not universally required to notify courts or opposing counsel simply because they used generative AI to assist with research or draft pleadings. However, individual judges retain the authority to issue local rules or specific case management orders mandating AI disclosure statements within their respective courtrooms.

Data security represents another major focus of the administrative notice. The guidance cautions attorneys that entering sensitive, proprietary, or privileged client data into public or unvetted commercial AI models risks waiving attorney-client privilege. Because many commercial generative systems store and process user inputs to train subsequent model iterations, practitioners must scrutinise vendor terms of service to maintain statutory confidentiality.

The directive emerges after several widely publicised cases in which attorneys submitted hallucinations produced by automated tools to state and federal courts. By declining to prohibit the technology outright, the Unified Court System is aligning its administrative posture with professional responsibility standards rather than technological bans.

Will the court system's reliance on voluntary verification and existing sanctions prove sufficient to deter fictitious case law from entering the official record?