Beginning in 2023 and accelerating through the hallucinated-citation sanctions era, individual federal judges adopted standing orders governing generative AI in filings — courtroom-specific rules requiring disclosure of AI assistance, certification of human review, or prohibitions on AI-drafted content entering orders unverified — and the population now runs to hundreds of courtrooms with materially varying texts. There is no circuit-wide rule, no uniform national standard, and no single registry: the operative law of AI-assisted filing is a patchwork of judges' preferences that litigators must inventory per judge, per case.
3G Times publishes information, not legal advice. Filing obligations are judge-specific and case-specific; this analysis addresses the operational discipline of tracking them.
What do the standing orders actually vary on?
Everything a drafter might assume is settled. Coverage: some orders reach only briefs and motions; others sweep declarations, discovery responses, and correspondence. Disclosure: some require affirmatively stating that generative AI was used and in what role; others require disclosure only on request, or only for jury materials. Certification: the most common pattern — counsel certifies that AI-assisted text was reviewed and verified by a human, with the certification's wording varying from signature-block implications to separate notations. Prohibitions: a minority of orders bar AI-drafted content outright, at least for specified document classes. Confidentiality riders: a growing subset restricts filing sealed or confidential material through public AI tools — a duty that tracks the privilege analysis bar guidance already imposes, now with a courtroom consequence. The variance is the point: the same filing that satisfies one judge's standing order violates the next judge's across the hallway.
| Order element | Range across courtrooms | Operational answer |
|---|---|---|
| Instruments covered | Briefs only ↔ all filings | Per-judge coverage flag |
| Disclosure trigger | Always ↔ on request ↔ never | Drafting-workflow gate |
| Certification form | Signature ↔ separate notation | Template library keyed to judge |
| Prohibitions | None ↔ AI-drafted bans | Escalation path to counsel |
| Confidentiality riders | Generic ↔ tool-specific | Sealed-material workflow gate |
Why hasn't the system standardized?
Because the incentives don't align yet. The judicial conference machinery moves deliberately; the orders' authors are responding to local facts — the hallucinated filings that landed in their districts — with local remedies; and the technology's role in drafting keeps shifting, making any uniform text dated on arrival. Proposed uniform rules and model orders circulate; none has achieved adoption at scale. Meanwhile the appellate and district clerks' websites host the orders unevenly — some judges publish in prominent places, others bury the rule in chambers pages or case-specific orders entered without fanfare — which is why experienced litigation-support teams maintain internal registries: the standing-order database, updated by docket-watchers, consulted at every case-opening and every new-judge event.
What does compliance look like as a workflow?
The mature pattern gates drafting, not filing-day heroics. At case opening: pull the standing orders for the assigned judge — and the magistrate's — into the case file, with the coverage flags translated into the drafting platform's checklists. During drafting: the AI-assistance log records which tools touched which documents, at what role — the record that makes any disclosure true. Before filing: the certification step matches the judge's required form, with the human-review attestation tied to the review that actually happened. And for sealed or confidential material: the confidentiality riders gate which tools were eligible at all. The workflow's cost is hours; its absence's cost is the violation that reads as misrepresentation — the sanctions era's lesson applied to a rule set that changes per courtroom.
How do the orders interact with ethics duties?
They floor them, they do not replace them. The bar's guidance — the ABA's Opinion 512 lineage and its state analogues — imposes competence, confidentiality, and candor duties regardless of courtroom; the standing orders add judge-specific procedural requirements on top. A filing that satisfies a lenient order still answers to the ethics rules; a filing that violates a strict order cannot be defended by pointing at ethics-compliant tool use. Litigation teams that read the two layers as one checklist — ethics defaults plus judge-specific riders — avoid the trap of treating either as sufficient alone.
What does this mean in practice?
- Maintain the registry — per-judge standing orders with coverage flags, owned by litigation support, updated on docket events and new appointments.
- Log AI assistance during drafting — the disclosure duty is only as honest as the record behind it, and memory is not a record.
- Gate sealed material at tool selection — the confidentiality riders and the privilege rules agree on this more than on anything else.
- Re-pull the full order set on any reassignment or transfer — the case that changes judges changes rules; the workflow's trigger list should name the event.
The standing-order patchwork is federal litigation's usual answer to a fast-moving practice question: local rules first, uniformity later. The teams that treat the patchwork as data — inventoried, flagged, gated into workflow — file anywhere without incident; the teams that assume yesterday's rule follows the case learn the variance from an order they did not want to be memorable.
A transition note for growing teams and litigation-support vendors: the registry works only as well as its maintenance triggers — new judges, published orders in later cases, and amendments to existing ones. Docket-watch automation covers the first two cheaply; the third requires periodic re-pulls, because the standing order that changed silently between filings is the one litigation teams never planned around.
The synthesis for litigation leaders: the patchwork is emphatically not a failure of the system but its first working draft — the same way e-filing rules, discovery protocols, and admission-of-technology orders all matured from courtroom-by-courtroom variance into standing frameworks. The teams that kept registries through these draft years will write the eventual uniform practice from their own checklists; the others will inherit whatever the committees produce.
Do magistrate judges issue them too?
Yes — and the discovery-context orders can be stricter than the district judge's, reaching meet-and-confer representations and discovery responses. The registry's per-case pull covers both benches: the workflow that remembered the presiding judge and forgot the magistrate learns the omission at discovery's worst moment.
Frequently asked questions
Where are the standing orders published?
Wherever each judge chose — court websites, chambers pages, case-management orders. There is no authoritative national index, which is precisely why experienced teams maintain internal registries and treat third-party compilations as starting points, not authority.
Do these orders apply to transactional work?
No — they are courtroom rules for filings. Transactional practice answers to ethics guidance and engagement terms instead, which is why the AI-disclosure question in deals is a contract clause, not a judge's standing order.
The AI-assistance log deserves the same custody as the registry: tool, role, documents touched, retained with the matter. When a judge asks for disclosure after filing, the log is the difference between a supplementary disclosure drafted in an afternoon and a reconstruction project under a show-cause deadline.
What happens if the order is missed?
Consequences range from corrective orders to disclosure demands to, in the worst cases, sanctions analysis — especially where an unverified filing also contained errors. The violation is procedural; the story it tells about verification is what makes it expensive.
For more context, read Verifying Authority in the AI Research Era: The Citator Discipline That Still Decides Cases.
For more context, read agentic ai legal supervision.
For more context, read outside counsel guidelines ai billing.