AI Prompts as Opinion Work Product: Tym v. Cerno

In April 2026, a federal court in New Mexico extended and consolidated the emerging judicial consensus: the prompts a litigant sends to ChatGPT or Claude, and the responses received, are protected work product when generated in the course of preparing filings or for hearings. The ruling in Tym v. Cerno applies the framework from Morgan v. V2X directly and confirms that AI interactions do not automatically forfeit their protected character merely because they involve a third-party platform.

Background and Facts

The plaintiff in Tym v. Cerno was a pro se litigant proceeding against Cerno, Flannery and the New Mexico Health Care Authority in the District of New Mexico. The case arose in a context that has become increasingly familiar in American courts: an unrepresented party using AI tools — specifically ChatGPT and Claude — both to prepare written filings and to assist during hearings.

The use of ChatGPT and Claude in hearing preparation is worth pausing on. It suggests that the plaintiff was not merely using AI to draft documents in advance, but was also consulting AI tools in real time or near-real time as part of hearing strategy. Whether the AI use extended to active consultation during the hearing itself is not entirely clear from the public record, but the court’s framing — “for hearings” — encompasses at minimum the preparation of arguments, anticipated questions, and strategic positions in anticipation of court proceedings.

The defendants moved to obtain discovery of the plaintiff’s AI interactions: the prompts submitted, the outputs received, and the extent of reliance on AI in preparing litigation materials. The legal theory underlying the motion was the by-now-familiar argument that AI tool use falls outside the scope of work-product protection, either because AI is a “third party” whose involvement waives the protection or because pro se litigants lack access to the doctrine in the first instance.

The court, adopting the reasoning of Morgan v. V2X explicitly, rejected both arguments.

What the Court Decided and Why

The court’s core holding in Tym v. Cerno is that AI interactions used to prepare filings or for hearings may be shielded as work product. The conditional framing — “may be shielded” — reflects the case-by-case nature of the inquiry rather than any ambiguity about the doctrine’s availability. Whether any specific interaction is protected will depend on whether its disclosure would reveal the litigant’s mental impressions, legal strategies, or analytical approach — the standard opinion work-product inquiry applied to a new category of material.

The court reasoned, as the Morgan court had before it, that the work-product doctrine protects the process of litigation preparation, not merely the documents that emerge from that process. A litigant who asks ChatGPT to help identify the strongest counter-argument to the opposing party’s position is engaged in exactly the kind of strategic analysis that work-product doctrine exists to protect. The prompt reflects the litigant’s assessment of the case; the response reflects the output of an analytical process initiated and directed by the litigant. Both are the product of anticipation of litigation, and both can therefore attract protection.

The extension of this reasoning to hearing preparation is significant. Courts and practitioners have sometimes drawn an implicit distinction between document preparation — where work-product thinking clearly applies — and in-hearing conduct, which is governed by different procedural rules. The Tym court’s reference to AI use “for hearings” as potentially protected suggests that the protection extends to the full arc of litigation preparation, including the final stages before a court appearance. This does not mean that AI use during a hearing is similarly protected — that question raises distinct concerns about disclosure obligations and courtroom procedure — but it means that a litigant’s AI-assisted preparation for a hearing is not a category apart from other protected preparatory work.

The defendants’ disclosure motion was denied. The court’s adoption of Morgan reasoning meant that the defendants would need to establish that something beyond the bare fact of AI use warranted discovery — a showing they had not made.

The Principle Established

The line of cases from Warner through Morgan to Tym has, by April 2026, established a reasonably coherent framework. First, AI interactions in litigation preparation are, as a matter of doctrine, capable of attracting work-product protection. Second, the protection is not automatic: a litigant who asserts it must be able to show that the specific materials sought would reveal mental impressions or strategy. Third, the bare identity of the AI tool used is likely not protected standing alone, but the content of interactions — prompts and outputs — directed at legal analysis and strategy is protected. Fourth, the protection is available to pro se litigants, not merely to represented parties.

Tym v. Cerno adds to this framework the specific confirmation that hearing preparation using AI falls within the protected category. It also reinforces the Morgan reasoning in a new jurisdiction, demonstrating that the framework is being adopted across multiple federal circuits and is not confined to the specific facts or district of any single ruling.

For litigants and their counsel, the practical implication is clear: maintain records of AI interactions that form part of litigation preparation, treat them as you would any other work-product material, and be prepared to assert the protection specifically and with particularity if discovery is sought. The protection is real, but it requires active invocation and substantiation.

The India Angle

Indian procedural law does not contain a direct analogue to the American work-product doctrine codified in FRCP 26(b)(3). However, the underlying concept — that a lawyer’s mental processes and case strategy should be shielded from adversarial discovery — is recognised in Indian law through related doctrines. Section 129 of the Indian Evidence Act, 1872 protects confidential communications between a client and attorney from compelled disclosure; Section 126 prohibits an advocate from disclosing professional communications without consent. The Indian courts have also, in various contexts, recognised the sanctity of the lawyer’s preparatory work.

The question of how these provisions apply to AI-assisted preparation has not yet been tested in Indian litigation, but the Tym framework offers Indian advocates a principled basis for asserting protection over their AI interactions. More immediately relevant is the hearing-preparation dimension of the case. Indian advocates who use AI tools to prepare arguments, anticipate questions from the bench, or develop hearing strategy are engaged in the same kind of protected preparatory work that the Tym court shielded. If discovery of such interactions were ever sought in Indian proceedings — a scenario that is not yet common but is no longer inconceivable as AI use becomes widespread — the reasoning of Warner, Morgan, and Tym collectively provides a substantial doctrinal foundation for resisting such disclosure.

Case Details