The proposal follows several fabricated-citation problems in the circuit. In Lichfield v. Kubler (July 27, 2026), the court held that a brief citing a nonexistent authority violated the existing reasonable-inquiry certification but imposed no sanctions. It assumed, without finding, that generative AI caused the errors. In June, a self-represented litigant's filings quoted language that did not appear in the cited cases, and the panel warned that future hallucinated citations could draw sanctions. In Picon-Diaz v. Bondi (February 2026), Judge Paul Kelly wrote that a citation and its quotations appeared fabricated, likely from unreviewed AI output. Details of how the proposal will treat sanctions, and what standards will govern the reviewer's confirmation, are not yet settled.
For attorneys practicing before the Tenth Circuit, the practical change is that the signature on a filing now carries a verification duty for AI-assisted work, with no disclosure step to satisfy. Watch the final text in December for how "reviewed" and "accurate" are defined and what sanctions apply. The proposal differs from the Fifth Circuit's 2023 AI disclosure certification, which was dropped in 2024 after commenters argued existing rules already covered accuracy. Commentary published October 23 urges the court to adopt the rule with clearer standards and proportionate enforcement. If adopted, the Tenth Circuit would be the first federal appeals court to require human verification as part of its AI rules, and observers expect other circuits to watch closely. Florida's June 2026 citation-accuracy certification, which does not mention AI, and Connecticut's Practice Book § 4-9, effective June 23, 2026, point the same direction.