Under Fed. R. Civ. P. 83(b), judges have authority to 'regulate practice in any manner consistent with federal law, . . . and the district’s local rules.' Utilizing Rule 83(b), courts around the United States have crafted different types of orders addressing the use of AI by attorneys. As of May 1, 2026, there are over 100 active orders issued by federal courts related to the use of GenAI by litigants. Many of these orders were issued by individual judges, others on behalf of an entire court, and some standing orders are now integrated into local court rules. 

While as of July 2026, there are over 100 AI-specific orders issued by federal judges, most judges do not have such orders. This could be because they deem the existing rules governing attorney practice sufficient or simply have not considered the issue.

Judges may issue a standing order that prohibits counsel from using AI to produce federal court filings. However, a broad prohibition may inadvertently include permissible uses of AI, like a Westlaw Boolean search or the use of Microsoft Word’s spellcheck. Courts choosing to issue broad prohibitions should consider narrowing the scope of their orders to generative AI.

There are different types of general disclosure orders. Some require filers to state that they ‘used’ AI; others also require disclosure of the specific GenAI tool that was employed, like ChatGPT. These rules may require the filer to ‘notify’ the court of AI use on the face of the relevant document or by submitting a separate certification in the certificate of compliance. General disclosure orders typically include a warning that failure to comply will result in sanctions under the Federal Rules of Civil Procedure. 

Verification orders require all filers to submit a certification stating either that GenAI was not (and will not be) used to prepare a court filing or that GenAI was used and all content was reviewed by a human for accuracy.

Most heightened disclosure orders include three requirements: (1) disclosure that AI was used to prepare a filing and/or the particular AI tool used; (2) identification of specific portions of the filing that were generated by AI; and (3) a certification that a human being verified the AI-generated materials, including citations.

A ‘warning’ or ‘sanction reminder’ is a standing order stating that, while AI use is not limited or prohibited, misuse of the technology may subject the filing party to sanctions under the Federal Rules of Civil Procedure and/or the disciplinary scheme of the state’s Rules of Professional Conduct. As of April 2026, this type of order is the one most commonly used in the federal courts.

Some judges have addressed confidentiality in their standing orders, requiring filers to certify that GenAI use has not resulted in the unauthorized disclosure of confidential information to a third party. Judges are now confronting—and disagreeing about—confidentiality expectations in the context of GenAI. Compare United States v. Heppner, 2026 WL 436479, at *2-4 (S.D.N.Y. Feb. 17, 2026) (finding attorney-client and work-product privileges inapplicable when a counseled defendant communicated with a GenAI tool) with Warner v. Gilbarco, Inc., 2026 WL 373043, at *4 (E.D. Mich. Feb. 10, 2026) (finding pro se litigant’s communications with a GenAI tool protected under the work-product doctrine).

Another strategy commentators have discussed is a standing order mandating that citations to all cases, statutes, and regulations in a filing be hyperlinked to a trusted legal database such as Westlaw, Lexis, Bloomberg Law, or an official court/government website. The judiciary’s case management software is capable of imposing such a requirement, but this type of rule has not yet been adopted. Drawbacks to such a requirement include GenAI’s ability to create a fake hyperlink; access to justice concerns, as not every filer has access to fee-based legal databases that can produce such hyperlinks; and the potential for a hyperlink rule to conflict with a district court’s local rules.