US Appeals Court Rule Would Require Lawyers To Certify AI-prepared Filings

The growing use of generative artificial intelligence in legal practice is forcing courts to confront a fundamental question: who remains responsible when technology helps prepare a legal filing that contains a serious error?

 

The US Court of Appeals for the Tenth Circuit has proposed a rule that would place the answer firmly on the human who files the document. If adopted, the rule would require lawyers and self-represented litigants to certify that court documents prepared with the assistance of generative AI have been reviewed by a human for accuracy, legal compliance and ethical obligations.

 

The proposal, announced on September 18, 2026, could mark a significant development in the regulation of AI-assisted advocacy. The Tenth Circuit would become the first of the 13 US federal appeals courts to adopt a rule specifically addressing the use of generative AI in court filings, according to Reuters. The proposed changes are intended to take effect on January 1, 2027, subject to the outcome of the consultation process.

 

From AI Assistance To Human Accountability

 

The proposed approach does not seek to prohibit lawyers from using generative AI. Instead, it focuses on what happens after AI has been used.

 

Under the proposal, a person filing an AI-assisted document would have to certify that a human independently verified the existence of cited authorities, the accuracy of quotations and citations, the legal analysis and compliance with applicable rules and ethical requirements.

 

That distinction is important. The issue is not simply whether AI can draft a brief, summarise a judgment or help formulate legal arguments. The central concern is whether a lawyer can responsibly place the resulting document before a court without independently checking its contents.

 

Generative AI systems can produce fluent and convincing legal prose while generating incorrect authorities or attributing propositions of law to cases that do not support them. Courts in the US have already encountered filings containing fabricated or inaccurate citations, prompting sanctions under existing procedural and professional rules. The Tenth Circuit's proposal effectively seeks to make the human verification obligation more explicit rather than treating AI-generated errors as an unforeseen technological problem.

 

Existing Duties Are Not Disappearing

 

The proposal also highlights an important legal principle: the arrival of AI does not remove existing responsibilities attached to signing and filing documents.

 

Chris Wolpert, clerk of the Tenth Circuit, told Reuters that the court continues to encounter situations in which existing requirements are not being followed in filings prepared with AI assistance. He said the proposed rule is intended to reinforce the representations that accompany the act of signing and filing a document, including representations concerning reasonable inquiry, accuracy and truthfulness.

 

In practical terms, this means that AI would remain a tool rather than a source of legal responsibility. A lawyer may delegate drafting assistance to technology, but the professional obligation attached to the final filing would remain with the lawyer.

 

This could have consequences for how law firms develop internal AI policies. Verification may need to become a documented part of the litigation workflow, particularly where AI has been used to identify authorities, summarise case law or generate substantive legal arguments.

 

Why Appellate Courts Face A Particular Risk

 

The proposed rule is especially significant because appellate litigation depends heavily on the accuracy of legal authorities and the reasoning connecting those authorities to the issues under review.

 

An erroneous citation in an appellate brief can do more than create an embarrassing drafting error. If a cited case does not exist, does not contain the quoted language or does not support the legal proposition for which it is offered, the court's ability to assess the argument can be undermined.

 

Existing rules already provide courts with mechanisms to respond. Lawyers whose filings contain fabricated authorities can face monetary sanctions, disciplinary referrals or dismissal of an appeal. The Tenth Circuit's proposal would add potential consequences including treating the affected issue as waived or striking a non-compliant filing from the docket.

 

The proposed framework therefore moves beyond the question of whether AI use is permissible. It addresses whether the use of AI changes the level of diligence expected from the person submitting the final document. The answer proposed by the Tenth Circuit is effectively that technological assistance does not reduce the obligation to check the work.

 

A Different Direction From The Fifth Circuit

 

The Tenth Circuit's proposal also illustrates the lack of a uniform approach across the federal appellate system.

 

The Fifth US Circuit Court of Appeals proposed an appellate AI rule in 2024 but abandoned the plan after members of the legal profession objected. Critics argued that existing rules were sufficient to address inaccurate or improper filings.

 

That earlier experience demonstrates the difficulty courts face in regulating a technology that is developing faster than traditional procedural rules.

 

One approach is to rely on broad existing duties governing candour, accuracy and professional conduct. Another is to introduce AI-specific requirements that make those responsibilities more visible to lawyers and litigants.

 

The Tenth Circuit proposal represents the latter approach. Its significance may ultimately extend beyond the six states within the circuit because other federal courts will be able to observe whether a dedicated AI rule improves compliance without creating unnecessary procedural burdens.

 

The Rule Would Cover Lawyers And Self-Represented Litigants

 

Another notable feature is the proposed application to both lawyers and self-represented litigants.

 

The Tenth Circuit has jurisdiction over federal appeals arising from Colorado, Kansas, Oklahoma, New Mexico, Utah and Wyoming. If implemented, the proposed certification requirement would apply to appellate cases across those states from January 1, 2027.

 

Including self-represented parties recognises that the risks associated with generative AI are not confined to professional legal practice. Individuals without legal representation increasingly have access to AI systems capable of producing sophisticated-looking legal documents.

 

For courts, however, the challenge may be greater when a litigant has little understanding of legal research or the limitations of AI-generated material. A requirement to independently verify authorities and legal analysis could therefore place a significant responsibility on users who may have turned to AI precisely because they lack legal expertise.

 

The Emerging Meaning Of Human Review

 

The most important aspect of the proposed rule may be its definition of what meaningful human oversight should involve.

 

Simply reading an AI-generated brief for obvious grammatical or factual errors would not appear to satisfy the proposed standard. The human reviewer would be expected to independently verify authorities, quotations, citations and legal analysis. That suggests a substantive review rather than a superficial check.

 

This distinction could become increasingly important as AI systems improve. More convincing output may make errors harder, rather than easier, to detect. A document that reads like a professionally researched brief can create a false impression of reliability if its underlying authorities have not been checked.

 

The proposed rule therefore treats human involvement as an accountability mechanism rather than merely a final editing stage.

 

A Possible Model For AI Governance In Legal Practice

 

The Tenth Circuit's initiative could also influence how courts and law firms approach AI governance more broadly.

 

Instead of attempting to regulate every type of AI use, courts may increasingly focus on the point at which technology-generated material enters the formal legal process. The proposed certification requirement effectively creates a boundary: lawyers may use technology in preparing their work, but they remain responsible for what they ultimately present to the court.

 

That approach could prove more adaptable than attempting to regulate specific AI products or technologies. As generative AI tools evolve, the underlying obligation to verify legal work would remain largely unchanged.

 

The proposal also reflects a broader shift in legal practice. AI is moving from an experimental tool to part of everyday professional workflows, while courts are simultaneously confronting the consequences of unreliable output. The emerging regulatory question is consequently less about whether lawyers should use AI and more about the controls that must accompany its use.

 

Public Consultation Underway

 

The Tenth Circuit is accepting public comments on the proposed rule until October 18, 2026, and plans to finalise its revised rules by December 1. If adopted, the new provisions would take effect on January 1, 2027.

 

The consultation will provide lawyers and other interested parties an opportunity to assess whether a specific AI certification requirement is necessary, how it should operate and whether the proposed sanctions are proportionate.

 

Whatever the final form of the rule, the underlying message is already clear: using artificial intelligence to prepare a legal filing does not transfer responsibility for that filing to the technology. As courts develop their response to generative AI, human verification is emerging as a central principle of legal accountability.