Judge Brantley Starr's AI Standing Order: The Template Every Court Copied

Part of: Court AI Standing Orders by Jurisdiction

Why the Starr order matters more than its individual docket

Judge Starr’s May 2023 standing order predates almost every other AI-specific court rule discussed anywhere in this cluster. At the time it issued, generative AI tools were roughly six months into mainstream public availability, and no other federal judge had yet built a standing procedural requirement around their use in filings. That timing is the reason the order has become the reference point the rest of the field is measured against: it wasn’t written in response to an established body of prior court orders to draw on, it was the first one (AI Vortex, “Starr AI Standing Order: What Federal Courts Are Copying”).

What the order actually requires

The mechanics are straightforward and worth stating precisely, because “AI standing order” gets used loosely to describe several structurally different approaches (see the pillar’s Type 1 through Type 4 breakdown). Judge Starr’s order requires every attorney appearing before him to file a certificate attesting to one of two things: that no portion of the filing was drafted by generative AI, or that any AI-drafted language was checked for accuracy by a human being — specifically using print reporters or traditional legal databases, not relying on the AI tool’s own internal citation verification. A filing submitted without this certificate is stricken.

The verification method specified is the detail most secondary summaries compress or drop: the order doesn’t just ask attorneys to “check” AI output in the abstract, it specifies checking against print reporters or traditional legal databases — a standard that implicitly treats an AI tool’s own citation-checking feature (if it has one) as insufficient on its own, since the whole premise of the order is that AI tools can generate citations that look correct but aren’t (AI Vortex, “Starr AI Standing Order”).

What later orders visibly borrowed

The clearest evidence that the Starr order became a template, rather than a one-off, is in how closely later orders track its structure. Judge Nina Y. Wang’s District of Colorado order, effective December 1, 2025, requires every filing to contain a signed AI Certification from every individual who contributed to drafting, affirming either no generative AI use or 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” — language that maps directly onto the two-branch certify-or-disclaim structure Judge Starr’s order pioneered, with more explicit procedural detail (model certification language, an explicit consequence — “may be stricken without substantive consideration and with leave to re-file a compliant document”) layered on top (full text at cod.uscourts.gov, Wang Chambers, effective Dec. 1, 2025).

The pattern that emerges from comparing the two orders directly: Judge Starr established the core mechanism (certify or disclaim, striking as the consequence for non-compliance). Judge Wang’s order, more than two years later, kept that mechanism intact while adding procedural refinements — model language, an explicit re-filing path, and an explicit tie-back to Federal Rule of Civil Procedure 11 and 28 U.S.C. § 1927 as obligations that continue regardless of the certification.

Why “certify, don’t ban” became the dominant model

The pillar’s review of the Ropes & Gray tracker shows roughly 161 entries requiring disclosure and/or verification against only 5 outright prohibiting AI use entirely — meaning the certify-don’t-ban structure Judge Starr’s order established is, by a wide margin, the dominant model among courts that have addressed AI use with a hard requirement at all (Ropes & Gray AI Court Order Tracker). That imbalance is a reasonable proxy for which model other judges found more workable: a certification requirement lets a court avoid policing which specific tools attorneys use, while still creating a sanctionable, bright-line failure point — a missing or false certificate — that doesn’t require the court to first prove the underlying citations were fabricated before it can act.

What this means for anyone filing before a certification-model court

• Read the specific order’s verification-method language, not just whether a certificate is required. Judge Starr’s order specifies print reporters or traditional legal databases; other certify-model orders may phrase the verification standard differently, and the exact wording controls what counts as compliant.
• A missing certificate is a procedural failure independent of citation accuracy. A filing with entirely accurate citations can still be stricken if the required certificate isn’t attached — the certificate itself is the compliance gate, not a formality layered on top of substantively correct work.
• Newer orders add detail, not a different core model. Anyone already compliant with the Starr-style certify-or-disclaim structure is well-positioned for most later orders in this family, which tend to refine rather than replace the mechanism.
• The underlying Rule 11 and 28 U.S.C. § 1927 obligations exist regardless of any certification requirement. Judge Wang’s order says so explicitly; treat certification-model compliance as an addition to existing obligations, not a replacement for them.

Related reading

• Court AI Standing Orders by Jurisdiction — the pillar this spoke expands on.
• Inside N.D. Ohio’s Total AI Ban: Judge Boyko’s Standing Order — the contrasting Type 2 (outright prohibition) model.
• citationsafe.com/sanctions-database
• citationsafe.com/verify

Sources

• AI Vortex, “Starr AI Standing Order: What Federal Courts Are Copying”.
• 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.
• Ropes & Gray, AI Court Order Tracker, category filter counts accessed 2026-07-16.

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 →