Court AI Standing Orders by Jurisdiction: What Judges Actually Require Before You File

Published 2026-07-16 · By Andy Gaber, founder of Citation Safe

>

There is no “the” federal AI rule

Anyone searching for a single controlling rule on AI use in federal court filings will not find one. As of mid-2026, “there is not yet any single, district, or state-wide local rule on disclosing the use of AI in court filings; rather, individual judges have issued standing orders or individual rules” (Greenberg Traurig, Nov 2025). The Fifth Circuit itself, despite producing the single most influential individual order in this area (Judge Starr’s), has “considered AI disclosure rules but declined to adopt them” at the circuit level, relying instead on existing ethical duties and Federal Rule of Civil Procedure 11.

That absence of a uniform standard is not a gap waiting to be filled by a single forthcoming rule — it’s the structure itself. Compliance with “the AI rule” isn’t a single checklist item; it depends on which judge, which division, which district, and in some states, which specific court within that state, is going to see the filing.

How many orders actually exist, and what they require

Ropes & Gray’s publicly available tracker of AI-related standing orders, local rules, and court decisions gives the clearest available count of how large this patchwork already is. Filtering the tracker’s own category counts (accessed 2026-07-16) shows roughly 756 entries tagged “Applies to AI Used for Filings/Drafting,” 161 tagged “Requires Disclosure and/or Verification,” 402 tagged “Suggests Cautious Use of AI” without a hard requirement, 169 tagged “Court-Imposed Consequences — Attorneys/Law Firms,” and only 5 tagged “Prohibits Use of AI” outright (Ropes & Gray AI Court Order Tracker). Read together, those numbers describe a landscape where an outright ban is rare, a hard disclosure-or-certification requirement is a meaningful minority, and the most common posture is a caution — a warning that AI use is subject to existing rules, without a new procedural gate attached.

These four categories are worth treating as a rough typology, because most individual orders fall cleanly into one of them.

Type 1: Certification / disclosure requirement

This is the model that dominates the “hard requirement” category, and it traces to a single, widely-copied original. Judge Brantley Starr of the Northern District of Texas issued a standing order in May 2023 — before most courts had addressed the issue at all — requiring every attorney to file a certificate attesting either that no portion of the filing was AI-drafted, or that any AI-drafted language was checked for accuracy by a human using print reporters or traditional legal databases. Failure to file the certificate results in the filing being stricken (AI Vortex, “Starr AI Standing Order: What Federal Courts Are Copying”).

Judge Nina Y. Wang’s District of Colorado standing order, effective December 1, 2025, follows the same structural approach with more procedural detail. It requires every filing to contain a signed AI Certification, from every individual who contributed to drafting, either affirming no generative AI was used or affirming that AI-drafted language “was personally reviewed by the filer or another human for accuracy” and that “all legal citations reference actual non-fictitious cases or cited authority.” The order provides model certification language for both cases and states plainly that non-compliant filings “may be stricken without substantive consideration and with leave to re-file a compliant document” (full text at cod.uscourts.gov, Wang Chambers, effective Dec. 1, 2025).

This model — certify, don’t ban — is what most courts that have addressed AI at all have converged on. It does not prohibit AI use; it makes silence about AI use, or a false certification, the sanctionable event.

Type 2: Outright prohibition

A smaller number of individual judges have gone further and simply banned AI from the filing-preparation process. Judge Christopher A. Boyko of the Northern District of Ohio’s standing order states plainly: “no attorney for a party, or a pro se party, may use Artificial Intelligence (’AI’) in the preparation of any filing submitted to the Court.” Violations can draw sanctions “including, inter alia, striking the pleading from the record, the imposition of economic sanctions or contempt, and dismissal of the lawsuit.” Notably, the order carves out legal and internet search engines — Westlaw, LexisNexis, Google, Bing — from the ban, targeting generative drafting and research tools specifically rather than any tool that happens to use AI somewhere in its infrastructure (full text at ohnd.uscourts.gov).

This is the rarer model — the Ropes & Gray tracker’s count of 5 entries tagged “Prohibits Use of AI” outright, against 756 that apply to AI use in some fashion, suggests outright bans remain the exception rather than the emerging norm.

Type 3: Court-wide local rules, not individual judge orders

The two models above are typically issued by a single judge for their own docket. A separate and arguably more consequential category is the court-wide or district-wide local rule, which binds every filing in that court regardless of which judge is assigned.

Texas has built out the most developed state-level example of this. The Northern District of Texas adopted Local Civil Rule 7.2(f), effective September 2, 2025, requiring that any brief prepared using generative AI carry a “Use of Generative Artificial Intelligence” disclosure header on its first page — a rule that applies court-wide, independent of Judge Starr’s earlier individual order covering his own docket. The Eastern District of Texas followed in December 2025 with General Order 25-07, amending Local Rule CV-11(g) to address generative AI directly (search summary via legalaigovernance.com and aivortex.io trackers, cross-referenced against WebSearch results, July 2026). Texas is described in that same tracking as having “the most developed federal court AI rule framework of any state, with binding local rules in three of four federal districts” — a materially different posture from most states, where individual-judge orders remain the only mechanism.

Bankruptcy courts have moved on their own track. The Southern District of California’s Bankruptcy Court issued General Order 210 on November 18, 2025, governing “the use of generative Artificial Intelligence (AI) for all pleadings, motions, and papers filed in Court proceedings,” effective January 1, 2026 — a rule that applies to every filing in that court, not just filings before one judge (casb.uscourts.gov, General Order 210 announcement).

Type 4: Statewide rules, and the case of New York

New York represents the largest single jurisdiction to move on this at the state level, and it illustrates why “statewide rule” doesn’t necessarily mean “uniform requirement.” 22 NYCRR Part 161 takes effect June 1, 2026, and establishes a system-wide policy for all New York Unified Court System courts — but the policy itself sets a default of no mandatory disclosure requirement, while authorizing individual courts to opt into a stricter Model Rule (Appendix A) under which a filer’s signature certifies the paper contains no fabricated AI content. The certification obligation therefore applies only in the specific courts that affirmatively adopt the Model Rule, not statewide by default (per reporting summarized from Greenberg Traurig’s November 2025 analysis and the New York court system’s own rule text). In practice, this means “New York has a statewide AI rule” is true and largely uninformative on its own — the operative requirement still depends on whether the specific court a filing lands in has opted into the stricter tier.

Federal courts with no order at all

The single largest category in the Ropes & Gray tracker’s implicit denominator isn’t any of the four types above — it’s silence. Most federal judges, in most districts, have not issued any AI-specific standing order at all. For filings before those judges, the operative rules are the ones that predate generative AI entirely: Federal Rule of Civil Procedure 11’s requirement that factual and legal contentions be well-grounded, 28 U.S.C. § 1927’s sanctions for vexatious multiplication of proceedings, and each state’s ordinary attorney-conduct rules. Every standing order discussed above is explicit that it operates on top of these existing obligations, not in place of them — Judge Wang’s order states plainly that “all filings remain subject to Federal Rule of Civil Procedure 11, 28 U.S.C. § 1927, and applicable ethical rules” regardless of whether an AI certification was filed.

This is the point easy to lose in a jurisdiction-by-jurisdiction breakdown: the absence of a standing order is not the absence of risk. Every sanctioned case in Citation Safe’s tracked 49-case sanctions database was sanctioned under a court’s inherent authority, Rule 11, or an equivalent state rule — not, in most instances, because a specific AI standing order was violated. A standing order changes the procedural trigger (a missing certificate can get a filing struck before anyone even evaluates the citations inside it); it does not create the underlying obligation to file accurate documents, which existed before generative AI and applies whether or not a court has said anything about AI specifically.

What this means for a filing workflow across jurisdictions

• Check the specific judge’s standing orders before every filing, not just the district’s local rules. The dominant pattern (Starr, Wang) is individual-judge orders layered on top of, not replacing, district-wide rules — and a district with no AI-specific local rule can still have individual judges with strict certification requirements.
• A missing or non-compliant certification can get a filing struck on its own, independent of whether the underlying citations are accurate. In jurisdictions with a Type 1 or Type 3 requirement, the certification itself is a compliance gate, not paperwork.
• Silence is not safety. The large majority of federal judges have issued no AI-specific order, but every one of them retains authority under Rule 11 and inherent court power — the same authority that produced every sanction in the tracked 49-case database.
• New rules are still arriving. New York’s Part 161 takes effect June 1, 2026; Texas’s district-level rules only fully formed in the second half of 2025; this is an active area, not a settled one, and a jurisdiction’s posture at the start of 2026 should not be assumed current without checking.

Sources

• Greenberg Traurig, “Navigating AI Disclosure Rules in New York Courts”, November 2025.
• Ropes & Gray, AI Court Order Tracker, category filter counts accessed 2026-07-16.
• Judge Nina Y. Wang, Standing Order Regarding the Use of Generative Artificial Intelligence in Court Filings, D. Colorado, effective December 1, 2025 — full text fetched and quoted directly.
• Judge Christopher A. Boyko, Court’s Standing Order on the Use of Generative AI, N.D. Ohio — full text fetched and quoted directly.
• Southern District of California Bankruptcy Court, General Order 210: Filings Using Generative Artificial Intelligence, issued November 18, 2025, effective January 1, 2026.
• AI Vortex, “Starr AI Standing Order: What Federal Courts Are Copying” and “Texas AI Disclosure Rules for Court Filings (2026)” — secondary-source summaries of N.D. Tex. Local Civil Rule 7.2(f) and E.D. Tex. General Order 25-07; primary rule text not independently fetched this pass (flagged below).
• Citation Safe, Public AI Legal Sanction Case Database, for the cross-reference to the 49-case sanctions set.

Sourcing note for Phase B: the Texas district-level rule citations (N.D. Tex. Local Civil Rule 7.2(f), E.D. Tex. General Order 25-07) and the New York 22 NYCRR Part 161 Model Rule mechanics are sourced to law-firm-blog secondary summaries in this draft, not to primary court-issued PDF text fetched directly, unlike the Wang and Boyko orders which were fetched and quoted from primary source PDFs. Before Phase B commits this content, the primary rule text for the Texas and New York items should be pulled directly (uscourts.gov / nycourts.gov) to bring their sourcing up to the same standard as the rest of the piece.

Citation Safe is a verification workflow tool, not legal advice. No attorney-client relationship is created. Human review is always required. Court rules change; always confirm current requirements against the specific court’s published rules before filing.

—

Check a brief before you file it →