CHP 1010 McDowell, LLC v. Turpen (In re Escobedo)

What sanction did the court impose?

Bankruptcy Judge Michael E. Romero granted the motion in part and denied it in part on the merits (denied as to claims 1-5 and 7; granted as to claim 6 for fraud or defalcation while acting in a fiduciary capacity, with leave to amend). On the AI conduct, the Court did not impose monetary or evidentiary sanctions, but issued a formal warning in a separate section labeled 'D. Warning to Turpen Concerning Artificial Intelligence.' The Court held that Turpen's pro se status does not relax his obligations under Fed. R. Bankr. P. 9011, and that 'the use of artificial intelligence for research or drafting of pleadings, whether by attorneys or pro se parties, is subject to Rule 9011 compliance, and the failure to ensure a citation exists and stands for the proposition for which it is cited may be sanctionable.' Citing Coomer v. Lindell, 2025 WL 1865282, at *3 and *8 (D. Colo. Jul. 7, 2025), as the authority for the sanctionability framework.

Why does CHP 1010 McDowell, LLC v. Turpen (In re Escobedo) matter for law firms using AI?

The smoking-gun feature of this case is the AI-prompt artifact, not the fake citation. The Court’s footnote 45 quotes the surviving AI-instruction text directly from Turpen’s motion, where the text reads as a literal continuation of an AI prompt-and-response interaction: ‘If we assume arguendo (for the sake of argument) that you did act in a fiduciary capacity, the strategy should shift away from challenging the existence of a fiduciary duty and instead attack the lack of defalcation, embezzlement, or larceny. Here’s a revised version of that paragraph with this shift in emphasis: Revised § 523(a)(4) Argument (Assuming Fiduciary Capacity).’ Turpen evidently copied an AI tool’s response wholesale into his motion without removing the framing language. For partners and trustees thinking about how to flag AI conduct in adversary proceedings, this is a usable template: the ‘Here’s a revised version’ phrasing is a textbook generative AI fingerprint that survives copy-paste and is searchable on a docket.

The doctrinal move is the application of Fed. R. Bankr. P. 9011 to pro se parties using AI. Rule 9011 mirrors Fed. R. Civ. P. 11 in substance but applies in bankruptcy proceedings, including adversary proceedings filed under § 523 nondischargeability claims. Romero’s order makes clear that the rule applies equally to attorneys and pro se parties, and that the substantive obligation runs to citation accuracy: ‘the failure to ensure a citation exists and stands for the proposition for which it is cited may be sanctionable.’ That formulation tracks the Coomer v. Lindell standard the Court cites, and aligns with the broader 2025 federal pattern of treating pro se status as procedurally relevant but not as immunizing from sanctions.

The sanctions calibration is consistent with the warning-first pattern emerging in pro se AI cases: Romero declined to impose monetary sanctions on this first occurrence, instead docketing a formal warning that establishes the record for any subsequent sanctions analysis. The warning’s placement in a separately-labeled section (‘D. Warning to Turpen Concerning Artificial Intelligence’) is intentional formatting that makes the AI piece searchable and citable as a standalone holding, rather than being buried in a sanctions footnote. Bankruptcy practitioners reading this case as precedent should expect to cite Section D specifically rather than the order as a whole.