**Judge John G

Judge John G. Koeltl has served on the Southern District of New York bench since 1994, making him one of the most experienced federal judges in the country. A senior judge handling complex commercial litigation, securities cases, and civil rights matters, he operates in the S.D.N.Y.—a district that has become ground zero for AI-in-litigation controversies after the infamous Mata v. Avianca ChatGPT citation debacle unfolded in his courthouse.

The S.D.N.Y. doesn't have a district-wide AI disclosure rule. Instead, individual judges have issued their own standing orders and practice requirements. Whether Judge Koeltl has a specific AI order or not, he sits in a courthouse where every judge watched a colleague deal with fabricated AI citations in real time. The institutional memory is fresh, and the tolerance is low.

The S.D.N.Y. After Mata v. Avianca

The Southern District of New York became the epicenter of the AI-in-litigation crisis in June 2023 when Judge P. Kevin Castel sanctioned attorneys Steven Schwartz and Peter LoDuca for submitting a brief filled with fabricated case citations generated by ChatGPT in Mata v. Avianca. The sanctions—including a $5,000 fine—sent shockwaves through the legal profession and prompted judges nationwide to issue AI standing orders. For Judge Koeltl and every other S.D.N.Y. judge, Mata v. Avianca wasn't an abstract news story. It happened in their courthouse. That proximity has shaped how the entire district approaches AI-assisted filings, creating an environment where heightened scrutiny is the default.

S.D.N.Y. Individual Judge Requirements

As of 2025, the S.D.N.Y. has no single district-wide AI disclosure rule. Instead, six or more individual judges have issued standing orders with varying requirements—some mandate disclosure, others require certification that AI-generated content was verified, and some require both. The Greenberg Traurig analysis of New York court AI rules identifies specific S.D.N.Y. judges with individual practices addressing AI. This judge-by-judge approach means practitioners must check the assigned judge's standing orders for every single case. Don't assume that one S.D.N.Y. judge's requirements mirror another's—the variance is significant.

Complex Litigation and AI Risk in Judge Koeltl's Courtroom

Judge Koeltl's docket includes securities litigation, antitrust, complex commercial disputes, and civil rights cases—areas where legal analysis requires precision and where opposing counsel is typically sophisticated. In complex litigation, briefs cite dozens or hundreds of cases. The more citations in a filing, the higher the probability that an AI tool will hallucinate at least one. Complex litigation also involves specialized legal doctrines that AI tools handle poorly—securities law loss causation analysis, antitrust market definition, and multi-factor constitutional tests are exactly the areas where AI generates plausible-sounding but legally incorrect analysis.

The S.D.N.Y. AI Privilege Ruling

In February 2026, S.D.N.Y. Judge Jed Rakoff ruled that materials generated through a consumer AI tool at the prompting of a criminal defendant were not protected by attorney-client privilege or work product doctrine. This ruling—the first of its kind in the S.D.N.Y.—has implications beyond privilege. It signals that the court views AI interactions as fundamentally different from confidential attorney communications. For practitioners in Judge Koeltl's courtroom, this means AI-generated work product may be discoverable, and relying on AI without understanding the privilege implications adds another layer of risk.

Compliance Strategy for Judge Koeltl's Cases

Step 1: Check Judge Koeltl's individual rules and standing orders on the S.D.N.Y. website before filing. Step 2: Verify every citation independently through Westlaw or Lexis—in the district where Mata v. Avianca happened, this isn't optional. Step 3: Consider voluntary AI disclosure even if not required—the S.D.N.Y.'s institutional experience with AI failures means the court will react more severely to undisclosed AI problems. Step 4: Be aware that AI-generated documents may not be privileged under the Rakoff ruling. Step 5: In complex cases with many citations, build extra time into your filing schedule for verification—rushing AI-assisted complex litigation briefs is how Mata v. Avianca happened.

The Bottom Line: Before filing in Judge Koeltl's courtroom, verify every citation through traditional research tools, check his individual practice requirements, and remember that the S.D.N.Y. is the district where the AI citation crisis started. The institutional tolerance for AI errors is as low as it gets.