AI Disclosure in Court Filings: A Jurisdiction Tracker
Over 200 federal judges have standing orders on AI. Requirements range from certification to outright prohibition. Check before you file.
The patchwork of AI disclosure requirements across American courts is, frankly, a mess. There is no uniform standard, no central register, and no guarantee that the rules that applied last month still apply today. Judges issue, amend, and occasionally withdraw standing orders without coordinated timing or consistent format.
For practitioners, this creates a compliance environment that requires active monitoring, not passive awareness. You cannot assume. You must verify. Every time.
The Disclosure Landscape
Since Judge Brantley Starr of the Northern District of Texas issued the first AI-related standing order in 2023, hundreds of state and federal judges have followed. The approaches vary significantly.
The Certification Model. Some judges require attorneys to certify that either no portion of their filing relied on generative AI, or that any AI-drafted language was independently verified. Judge Starr pioneered this approach. It places an affirmative obligation on the attorney to make a statement — true or false — about AI use. The virtue is clarity. The risk is that a false certification creates an independent basis for sanctions.
The Disclosure-Plus-Verification Model. Judges in the Eastern District of Pennsylvania, including Judges Baylson and Pratter, require disclosure of AI use in filing preparation plus certification that each citation is accurate. This model does not prohibit AI — it requires transparency about its use and diligence about its output.
The Reminder Model. Some judges, like Judge Iain D. Johnston in the Northern District of Illinois, remind attorneys that existing duties under Federal Rules of Civil Procedure 11(b) and 26(g) already require accuracy in filings. This model says: we do not need new AI rules because the existing rules cover it. You are already obligated to verify everything you file, regardless of how it was produced.
The Prohibition Model. The Western District of North Carolina issued a standing order requiring lawyers to certify that AI was not used to prepare the brief. This is the most restrictive approach and raises questions about whether it will survive as AI becomes embedded in standard legal research tools.
The Enforcement Reality
Standing orders are not theoretical. They are enforced.
In August 2025, Judge Karoline Mehalchick of the Middle District of Pennsylvania sanctioned a pro se plaintiff for citing a non-existent case hallucinated by AI. The court found two violations: Rule 11(b) (legal contentions must be warranted) and the judge's own standing order requiring disclosure of AI use.
This is the pattern I expect to see repeatedly. AI disclosure requirements create a second basis for sanctions independent of the underlying accuracy obligation. An attorney who files a fabricated AI citation in a jurisdiction without a standing order faces Rule 11 exposure. An attorney who does the same in a jurisdiction with a disclosure requirement faces Rule 11 exposure plus a standing order violation.
The practical effect: disclosure requirements increase the sanctions surface area.
Notable Retreats
Not all standing orders have survived. Illinois Magistrate Judge Gabriel A. Fuentes withdrew his standing order approximately one year after issuing it, concluding it was "no longer necessary and slightly burdensome."
This is worth noting but should not be over-interpreted. One withdrawal does not signal a trend. The overwhelming direction is toward more disclosure requirements, not fewer. And Judge Fuentes's rationale — that existing professional responsibility rules adequately address AI use — does not reduce the underlying obligations. It simply returns enforcement to the traditional framework.
Pennsylvania: A Detailed Case Study
Pennsylvania provides the most comprehensive example of multi-layered AI requirements.
At the state level, Pennsylvania mandates explicit disclosure of AI use in all court submissions as of August 2024. This is not a local rule — it is a statewide filing requirement.
Joint Formal Opinion 2024-200, issued by the Pennsylvania and Philadelphia Bar Associations, addresses the ethical dimensions comprehensively: competence (knowing AI limitations), confidentiality (managing data exposure), communication (client disclosure), candor (truthfulness of AI content), and supervisory responsibilities (oversight of AI-using lawyers and staff).
The Pennsylvania Supreme Court issued guidance in September 2025 permitting judiciary personnel to use AI within defined parameters — acknowledging that the technology is acceptable when properly governed.
At the federal level, individual judges in the Middle and Eastern Districts have their own standing orders with specific requirements.
For a lawyer filing in Pennsylvania, the compliance stack includes: statewide disclosure mandate, bar association ethical guidance, and potentially judge-specific standing orders. All must be checked. All must be satisfied.
Voluntary Disclosure: The Strategic Question
In jurisdictions without mandatory disclosure requirements, voluntary disclosure is a strategic decision with legitimate arguments on both sides.
For disclosure: It demonstrates good faith and candor, may mitigate sanctions if problems are discovered later, aligns with emerging professional norms, and pre-empts attempts by opposing counsel to characterise AI use negatively.
Against disclosure: It may invite scrutiny where none was required, creates potential for inconsistent treatment across matters, adds administrative overhead, and may highlight AI use to judges who are sceptical of the technology.
My position: in most cases, disclose voluntarily. The risk of being caught without disclosure is greater than the risk of inviting scrutiny. And as disclosure requirements spread, early adoption of voluntary disclosure demonstrates a pattern of good faith that is valuable if questions arise later.
Practical Compliance
Before every filing:
- Check standing orders for the specific judge — not just the district, the individual judge
- Check court-wide local rules for AI provisions
- Document AI usage in your file — which tools, which portions of the filing, what verification was performed
- Verify every citation and factual assertion in AI-assisted content against primary sources
- Prepare appropriate disclosure language if required or if voluntary disclosure is your firm's standard
Tracking resources: The RAILS AI in Courts Tracker, the Ropes and Gray Court Order Tracker, and the Lexis Federal and State Court Rules Tracker are useful but not always current. Verify directly with the specific court when possible.
Documentation: Maintain records of which AI tools were used, what prompts were given, what output was generated, what verification was performed, and by whom. This documentation supports defence against sanctions if issues arise.
The Direction of Travel
More courts will impose disclosure requirements. More states will follow Pennsylvania's mandate. And as AI becomes embedded in standard legal tools — research platforms, document management systems, practice management software — the question of what constitutes "AI-assisted" will become increasingly complex.
The lawyers who build disclosure and verification practices now will be prepared for that evolution. Those who treat the current patchwork as a reason to delay will find themselves unprepared when the requirements arrive in their jurisdiction.
They will arrive. The only question is when.