AI Disclosure Rules in the District of New York, Southern

The Southern District of New York is the most important federal court in the country for AI disclosure law. Period. Covering Manhattan and the surrounding area, SDNY handles Wall Street securities litigation, Big Law commercial disputes, intellectual property battles, and the highest-stakes federal cases in American law. This is where Mata v. Avianca happened — the case that put AI hallucinations on every attorney's radar. Multiple SDNY judges have since issued standing orders on AI use, making this district the de facto national standard-setter for how lawyers interact with generative AI.

AI Disclosure Rules in the District of New York, Southern

The Southern District of New York does not have a single district-wide AI rule — but what it has is arguably more influential. Multiple individual judges have issued standing orders that collectively shape how AI is treated in federal litigation nationwide. The patchwork approach means that attorneys filing in SDNY need to check their specific judge's requirements, but the overall direction is unmistakable: disclosure, verification, and accountability.

Judge P. Kevin Castel issued one of the earliest federal standing orders on AI use, prompted directly by the fallout from Mata v. Avianca. His order requires attorneys to disclose AI use and certify accuracy. Judge John Cronan implemented a more detailed rule requiring attorneys to disclose AI use and certify they personally reviewed the filing for accuracy. Judge Arun Subramanian took a different approach, issuing a warning about generative AI pitfalls without mandating disclosure — a softer touch that still puts attorneys on notice.

This multi-judge, multi-approach framework is actually the most realistic model for how AI governance is evolving. The March 2026 NYC Bar study found that 41.7% of federal courts have no meaningful AI governance. SDNY is not in that group — it is at the opposite end of the spectrum, with the most developed AI jurisprudence of any district in the country.

The 2nd Circuit has not issued circuit-wide AI guidance, but with SDNY driving the conversation, it may not need to. The district's standing orders are effectively setting the floor for attorney conduct across the circuit.

Individual Judge Standing Orders

Three SDNY judges have issued particularly significant AI standing orders, each reflecting a different philosophy:

Judge P. Kevin Castel's order was a direct response to Mata v. Avianca and became one of the first federal AI standing orders in the country. It requires attorneys to disclose whether generative AI was used in preparing court filings and to certify that any AI-generated content was verified for accuracy. The order is straightforward: if you used AI, say so and confirm you checked the work.

Judge John Cronan's order goes further. Attorneys must disclose AI use and certify that they personally — not a paralegal, not an associate, not a junior attorney — reviewed the filing for accuracy. This personal certification requirement is significant because it places the obligation squarely on the signing attorney and eliminates the possibility of delegating AI verification down the chain.

Judge Arun Subramanian took a different path. Rather than mandating disclosure, he issued a cautionary warning about the pitfalls of generative AI without creating a formal requirement. This approach reflects the view that existing ethical rules — Rule 11, attorney competence standards — already cover AI risks without needing additional layers.

With over 300 federal judges nationwide now maintaining individual AI orders, SDNY's multi-judge approach has become a reference point for courts across the country. Attorneys practicing before any SDNY judge should assume that AI use will be scrutinized, whether or not that specific judge has a formal order.

Key AI Cases in SDNY

Mata v. Avianca, Inc. (SDNY, 2023) is the single most important case in the history of AI and litigation. Attorney Steven Schwartz used ChatGPT to draft an opposition brief and submitted it to the court with six case citations that were entirely fabricated — the cases did not exist. When opposing counsel and Judge Castel flagged the issue, Schwartz initially doubled down, asking ChatGPT to confirm the citations were real (it did, because that is what chatbots do). Judge Castel imposed $5,000 in sanctions on Schwartz and his colleague Peter LoDuca.

The dollar amount was almost beside the point. Mata v. Avianca became a global news story. It was cited in judicial AI orders across every circuit. It fundamentally changed how the legal profession talks about AI. Every subsequent AI standing order — from Texas to Hawaii to New Jersey — traces its lineage back to what happened in this courthouse.

The Couvrette sanctions case later escalated the financial stakes to $109,700, proving that Mata was not a one-off warning shot. For SDNY practitioners specifically, the message is unambiguous: this court has zero tolerance for unverified AI-generated content. The judges here were embarrassed by Mata, and they are not going to let it happen again.

SDNY's caseload makes this especially high-stakes. Securities fraud cases with billions in exposure. Patent disputes between major corporations. Antitrust actions that reshape entire industries. The accuracy bar was already the highest in federal practice. AI just raised it further.

What Attorneys in SDNY Should Do

**Check your assigned judge's AI standing order before every filing.** SDNY does not have a single rule — each judge may have different requirements. Judge Castel's order is different from Judge Cronan's, which is different from Judge Subramanian's approach. Pull the standing orders from PACER or the court website before you begin drafting.

**Build AI disclosure into your standard filing workflow.** In SDNY, assume disclosure is expected. Include a certification paragraph or footnote in every brief stating whether AI was used, which tool, and confirming human verification. This should be as routine as your signature block.

**Verify every citation twice.** SDNY opposing counsel will check your work. In this district, fabricated citations will be found. Pull every case on Westlaw or Lexis. Read the actual opinion. Confirm the holding matches your characterization. Do not rely on case summaries generated by any tool — AI or otherwise.

**Require personal attorney review, not delegation.** Judge Cronan's order specifically requires the signing attorney to personally verify accuracy. Even if your judge does not have that exact requirement, adopt this standard. Senior partners who sign AI-assisted briefs prepared by associates need to actually read and verify the content.

**Use enterprise-grade legal AI only — and document everything.** In the district that produced Mata v. Avianca, using free-tier ChatGPT for case research is indefensible. If your firm uses AI, invest in legal-specific platforms with citation verification features. Keep records of every AI interaction: prompts, outputs, and your verification steps.

The Bottom Line

The Southern District of New York is where AI met the federal judiciary, and the judiciary pushed back hard. Mata v. Avianca changed the profession. The standing orders from Judges Castel, Cronan, and Subramanian are the templates that courts across the country are adapting. If you practice in SDNY, you are practicing in the most scrutinized AI environment in federal law.

This is not a district where you figure out AI compliance later. Every filing, every citation, every representation needs to meet the standard that Mata established. The good news: if your workflow passes SDNY scrutiny, it passes scrutiny everywhere. Build your practice to this standard and you are covered in any federal court in the country.