Starting June 1, 2026, if you're a lawyer practicing in New York State courts and you've used AI to draft, research, or otherwise assist in preparing court filings, you're going to need to say so explicitly. The New York State Unified Court System has formally adopted a new rule mandating disclosure of artificial intelligence use in legal proceedings. No more quietly letting GPT-4 write your motion briefs while pretending it was you burning the midnight oil.

What the Rule Actually Does

The rule, adopted through the New York State Unified Court System and announced in coordination with the New York State Bar Association (NYSBA), requires attorneys to disclose when AI tools have been used in the preparation of documents submitted to the court. Think of it as a "this content was AI-assisted" label — but with legal teeth instead of a casual footnote.

This isn't just a nudge or a suggestion buried in a best-practices memo. It's a formal court rule. That means non-compliance isn't an awkward conversation with a ethics committee — it's a potential sanctions situation. The kind of thing that ends up in a disciplinary file.

Why Now? (And Why It Took This Long?)

Courts have been watching the legal profession's AI adoption with the nervous energy of someone who just handed a toddler a permanent marker. The concern isn't theoretical — there have already been high-profile cases where attorneys submitted briefs citing cases that simply do not exist, hallucinated wholesale by large language models that are, let's be clear, fundamentally incapable of "knowing" whether a citation is real.

The infamous Mata v. Avianca case from 2023 — where a lawyer submitted ChatGPT-generated fake case citations to a federal court — was essentially the legal world's "we need a rule for this" moment. New York is now formalizing what responsible practitioners should have been doing anyway: verifying AI output and being transparent about its use.

The Tradeoffs Nobody in the Press Release Mentions

Here's what the official announcement doesn't dwell on. Disclosure rules sound clean and simple until you start asking the hard questions. What counts as "AI use"? If you ran a contract through a grammar checker powered by a neural network, does that trigger disclosure? What about AI-assisted legal research platforms that have been embedded in Westlaw and LexisNexis for years? Where exactly is the line between "AI tool" and "software"?

These are not rhetorical questions. They're the exact ambiguities that will spawn a wave of follow-up guidance, bar opinions, and probably a few disciplinary proceedings before the dust settles. The rule sets a direction — the implementation details will be messy.

There's also the competitive dynamic to consider. Large firms with dedicated legal tech teams and in-house AI governance frameworks will adapt to this quickly. Solo practitioners and small firms — who arguably stand to benefit most from AI assistance — may find the compliance overhead disproportionately burdensome. That's a real tension worth watching.

What This Means If You're Building Legal AI Tools

If you're on the product side — building AI tools for legal workflows — this rule is a forcing function. Your platform now needs to generate a disclosure artifact that attorneys can actually use in filings. Audit trails, usage logs, and clear documentation of where AI was involved aren't just nice-to-haves anymore in New York. They're part of the compliance stack.

More broadly, expect other jurisdictions to follow. New York rarely moves alone on legal procedure, and with the federal courts still working through their own AI guidance, state-level rules like this one are going to multiply. Build your disclosure features now, not after your third enterprise customer in a different state asks why you don't have them.

The Bigger Picture

Look, AI in legal practice isn't going away. The efficiency gains in document review, contract analysis, and legal research are real and significant — we're talking about collapsing hours of work into minutes for certain task types. But "the model is useful" and "the model is reliable enough to submit to a court without supervision" are two very different claims, and courts are right to demand that attorneys own the distinction.

Mandatory disclosure doesn't hamper legitimate AI use. It just forces the humans in the loop to actually be in the loop — which, if your AI workflow didn't already include that step, was always the real problem.

New York's June 2026 deadline gives practitioners about a year to get their houses in order. That's enough time to build good habits. Whether the legal profession uses it that way is another question entirely.