Federal Courts Are Mandating AI Disclosure in Filings - What to Do Now

The Patchwork Is Becoming a Pattern

In 2023, a single federal judge in the Northern District of Texas issued a standing order requiring attorneys to disclose whether generative AI had been used in any filing. It was unusual. By 2026, similar rules - with meaningful variation - have spread across enough federal districts that "national" compliance is becoming a baseline operational concern for any firm appearing in federal court.

This article walks through the current state of federal AI disclosure rules, the three disclosure models that have emerged, what to put in your engagement letter, and the workflow change that makes compliance trivial.

For the broader ethics framework these court rules sit on top of, see AI ethics for lawyers: Rules 1.1, 1.6, and 5.3 and our companion piece on the latest state bar AI opinions.

The Three Disclosure Models

Almost every federal AI disclosure rule we have read falls into one of three categories. Understanding which category applies in a given district shapes the workflow.

Model 1: Certification of Verification

The most common. The rule requires the filing attorney to certify that any AI-generated content was independently verified - citations confirmed, statements of law checked against primary sources. Disclosure of AI use itself is not the trigger; the verification certification is.

Practically, this is the cleanest model. It does not require firms to publicly announce which tools they used. It just requires them to have actually done the verification work the Federal Rules already implicitly require.

Model 2: Affirmative Disclosure

A meaningful minority of districts require the attorney to affirmatively state, on the filing itself, whether generative AI was used in its preparation. Some require identifying the tool. Some require describing the type of use (research, drafting, summarization).

This model creates a workflow question more than an ethics question: what counts as "use"? Did spell-check count? Did Westlaw's AI count? Did the email auto-complete on Gmail count? Most rules clarify these are not what they mean - the trigger is substantive generative AI use in producing the content of the filing.

Model 3: Prohibition with Carve-Outs

A small number of courts have issued rules that effectively prohibit generative AI use in certain categories of filings unless specific safeguards are in place. These are the strictest. They are also the rarest - and increasingly out of step with the broader trend, which is toward verification certification rather than prohibition.

What Counts as "Use"

Across all three models, the consistent thread is this: the rules target substantive AI use in content production, not background productivity tools that happen to incorporate AI features.

What clearly triggers disclosure or verification rules:

• Asking a general AI tool to draft a brief, motion, or memo
• Asking an AI tool to research a legal question and produce the substantive analysis
• Using AI to generate a fact summary that appears in a filing
• Using AI to draft client communications attached as exhibits

What clearly does not trigger:

• Spell-check, grammar-check, predictive text
• Westlaw or Lexis search functionality that has always used AI under the hood
• Calendar suggestions, email triage, scheduling assistants
• AI features inside word processors that suggest formatting

The gray zone is in the middle: AI used to outline, AI used to suggest counter-arguments, AI used to compile fact patterns. Most rules treat these as substantive use requiring disclosure or verification certification. When in doubt, treat it as substantive.

A National Compliance Strategy

Even though no national rule exists, firms appearing in federal court regularly should adopt a national strategy. The simplest is what we recommend to every firm we onboard through our AI Operating System service:

1. Treat verification as universal

Every AI-assisted filing - regardless of jurisdiction - gets the same verification protocol: every citation confirmed, every statute pulled, every quotation matched to source. This satisfies Model 1 districts without thinking about jurisdiction.

2. Maintain an AI use record

A simple internal log: matter number, filing, AI tool used, type of use. Five seconds per entry. Makes Model 2 disclosures trivial when required. Useful for fee defense if billing is ever challenged. Documents the verification step for any future ethics inquiry.

3. Default to disclosure language for federal filings

Most firms we work with have moved to adding a standard verification certification to federal filings as a matter of practice, even where the rule does not technically require it. It is short, costs nothing, and inoculates against rule changes.

Sample language attorneys may adapt to their jurisdiction:

"Counsel certifies that any portion of this filing produced with the assistance of generative AI tools has been independently reviewed by undersigned counsel, that all case citations have been verified against primary sources, and that the filing complies with all applicable Federal Rules and local court rules."

Have your local counsel and ethics advisor approve specific language for your jurisdiction before using it.

4. Update engagement letters

Disclose AI use in engagement letters. We covered the language in should you bill AI time to clients. The disclosure obligations to clients and to courts are distinct but related - addressing both in your standard documents avoids inconsistency.

What to Do This Week

1. Audit federal districts where you appear regularly. Pull the standing orders from each judge or local rules from each district. Map which require disclosure, which require verification certification, and which are silent.

1. Update your filing checklist. Add a single line: "Federal filing? Confirm jurisdiction-specific AI disclosure rules satisfied." That one line catches most exposure.

1. Adopt a standard verification certification. Even in jurisdictions where the rule does not require it, the certification is harmless and forward-protective.

1. Train associates on the categories. Make sure every attorney on the team knows what counts as "use" and what does not. Confused junior attorneys are how mistakes happen.

1. Document everything. Verification checklists, AI use logs, training touchpoints. If a question is ever raised about a filing, you want a paper trail.

The Direction This Is Headed

The trajectory is clear. Federal courts are not going to standardize on a single national AI disclosure rule any time soon - the federal judiciary is structurally decentralized. What you will see instead is more districts adopting some version of verification certification, more standing orders from individual judges, and continued variation in the details.

Firms that build a single national compliance posture now - verification by default, disclosure language by default, AI use logging by default - will sail through any future rule changes. Firms that handle each new rule reactively will spend years playing whack-a-mole.

The work to set this up is modest. The protection is durable. Do it once and forget about it.

If you want help installing this across your firm, our AI Training program covers compliance workflows as part of every engagement. Or book a consult to walk through where your federal practice sits today.