The US Court of Appeals for the Tenth Circuit has proposed a rule expressly governing documents prepared with generative AI.
Under proposed Rule 46.5, signing or submitting a filing would certify that any AI-assisted document was reviewed by a human who independently verified every authority, quotation, citation and item of legal analysis, as well as compliance with applicable procedural and ethical rules. Breach could attract consequences including striking the document, waiver of issues, dismissal or affirmance of an appeal, monetary sanctions and disciplinary proceedings.
The proposal was released on 18 September for public comment and would take effect on 1 January 2027 if adopted. Reuters reports that it would be the first AI-specific rule introduced by one of the 13 US federal appellate courts.
The important shift is from general warnings about hallucinations to a formal representation made through the act of filing. The rule does not prohibit AI or require disclosure of the particular tool used. It instead makes human verification—and professional adoption of the resulting work—the enforceable boundary.
That approach is relevant in Australia, where courts including the Supreme Court of NSW already require practitioners to verify AI-assisted material. It suggests that durable legal-AI systems will need to produce evidence of review: verified sources, recorded corrections, unresolved issues and identifiable practitioner approval—not merely polished output.
Sources: Tenth Circuit proposal; Reuters.
Leave a Reply