In this article, Seed IP attorneys Jeffrey Danley, Jessica Gritton, and John Morgan report on a recent U.S. District Court decision holding that a pro se plaintiff’s use of a consumer-grade AI tool did not waive attorney work-product protection for the AI-related materials. This outcome differs from that in the recent Heppner case, holding that work-product protection did not apply to legal strategy materials generated by a criminal defendant using a consumer-grade AI tool. The authors explore these decisions and offer practical guidance for AI tool usage in patent prosecution and patent litigation.

‍Federal Courts Differ on AI Use and Waiver of Work-Product Protection

In a recent case before the U.S. District Court for the Eastern District of Michigan, Warner v. Gilbarco, Inc., discovery issues involving the use of a consumer-grade AI tool arose in which the pro se plaintiff had used generative AI (ChatGPT) to assist in drafting legal documents.  In the decision, the magistrate judge rejected defendant’s attempts to obtain the AI-related materials, holding that such materials were protected by the attorney work-product doctrine.  

The outcome in Warner differs from that in the also-recent United States v. Heppner, in which the New York district court judge found that legal strategy materials prepared using a different consumer-grade AI tool (Claude) by a client, without any involvement or direction by the attorney, were not protected by the work-product doctrine or the attorney-client privilege (see  our earlier article discussing Heppner).  

These two holdings, arriving at different decisions regarding waiver, illustrate the unsettled landscape of AI use and protections from discovery.

Warner Background

The defendant moved the court to require the pro se plaintiff to produce all case-related materials that the plaintiff had submitted to ChatGPT, as well as all documents generated by that tool.  According to the defendant, the plaintiff had uploaded case documents to ChatGPT, used ChatGPT to assist in drafting filings, and asked legal questions of ChatGPT.  The defendant sought production of all this material, or, alternatively, for the plaintiff to log this material as part of her privilege log.  The plaintiff asserted that this material was protected by the attorney work-product doctrine and that she used ChatGPT only as a drafting tool to assist her in drafting court filings.

The defendant advanced two arguments attacking the plaintiff’s assertion of protection.  First, the defendant argued that work product protection did not extend to pro se parties because the rule protects against the disclosure of “the mental impressions, conclusions, opinions, or legal theories” only of “a party’s attorney or other representative concerning the litigation.”  See Federal Rule of Civil Procedure (F.R.C.P.) 26(b)(3)(B).  According to the defendant, “the protection simply does not extend to a self-represented litigant.”  See dkt. No. 87 at 5.  Second, the defendant argued that ChatGPT was much more than a simple drafting tool, and instead evaluates uploaded data and performs detailed analysis in response to submitted prompts and questions.  According to the defendant, “[s]ubmitting queries to ChatGPT is analogous to Plaintiff calling up a third-party, non-spouse family member and discussing the lawsuit with him or her.”  Id. at 6.  In this situation, according to the defendant’s argument, “[i]f Plaintiff discussed ‘privileged’ matters with that family member, any such privilege would have been waived.”  Id.  

Warner Opinion

In a decision issued on February 3, 2026, magistrate judge Anthony Patti rejected both of the defendant’s arguments for waiver of the attorney work-product doctrine.  First, the court rejected the notion that the attorney work product doctrine did not extend to pro se litigants.  The court cited to F.R.C.P. 26(b)(3)(A), which applies more broadly than the provision cited by the defendant and protects against the discovery of “documents and tangible things that are prepared in anticipation of litigation or for trial by or for another party or its representative.”  See Decision, at 10.  The court further cited previous decisions in the same court extending the work product protection to pro se parties.  Id.  The court noted that the defendant’s request seeking production of the plaintiff’s drafts of legal documents effectively asked the court “to compel Plaintiff’s internal analysis and mental impressions—i.e., her thought process—rather than any existing document or evidence, which is not discoverable as a matter of law.”  Id. at 12.

The court rejected the defendant’s second argument for waiver by noting that the confidentiality requirements for the attorney work-product protection do not strictly forbid the disclosure of the protected material to any and all third parties.  Instead, according to the magistrate, the touchstone for waiver is whether the waiver is “to an adversary” or done “in a way likely to get in an adversary’s hand.”  See id. at 11.  Here, because ChatGPT is a tool and not a person, the magistrate found that there was no waiver of the work product protection.  The court, however, noted the differences between the requirements for finding waiver of attorney-client privilege (which may result from disclosure of materials to any third party) and for finding waiver of the work-product protection (which does not necessarily result from such a disclosure).

Practical Guidance

The decision in Warner illustrates nuances regarding the differences in applying the attorney-client privilege and the attorney work-product doctrine to materials submitted to and generated by consumer-grade generative AI tools.  These nuances should be kept in mind during both patent prosecution and patent litigation.

Depending upon how broadly the decision in Warner is interpreted, this portion of the decision is in tension with the decision made in Heppner, which found waiver of the attorney work-product protection when a represented client used a consumer-grade generative AI tool to seek information about his litigation.  Together, these decisions demonstrate that the law surrounding the use of generative AI is still in flux and may vary based on jurisdiction. According to Heppner, waiver of the attorney work product protection was appropriate under Second Circuit case law, which generally disfavors an expansive application of the protection and thus restricts it to protecting materials prepared by attorneys or their agents.  

The two decisions also demonstrate that courts may differ even in how they conceptualize generative AI. The court in Heppner appeared to treat the generative AI as a third party itself. In assessing whether privilege applied, the court stated that “what matters for the attorney-client privilege is whether [the defendant] intended to obtain legal advice from [the generative AI model] Claude” and cited Claude’s own responses as evidence that the model could not provide legal advice. By contrast, the court in Warner stated that generative AI “are tools, not persons, even if they may have administrators somewhere in the background.”

Clients should note that the decision in Warner (finding no waiver) likely does not directly apply in the patent prosecution setting because it addressed the attorney work product doctrine.  Notably, this doctrine applies only to materials generated in anticipation of litigation.  In most circumstances, patent prosecution occurs well before litigation is contemplated or anticipated, and therefore the work-product protection likely would not apply to patent prosecution materials submitted to or generated by commercially available, generative AI tools.  Instead, any protection of such materials from discovery would need to reside in the attorney-client privilege, which, as the magistrate in Warner noted, strictly forbids disclosure of privileged materials to third parties.

Accordingly, clients should not use consumer-grade generative AI tools without adequate confidentiality protections in any part of the invention disclosure or patent prosecution processes as it is possible that none of the correspondence with the AI tool will be privileged or protected.

In the litigation setting, Warner may provide some protection for the client’s use of documents submitted to or generated by a consumer-grade generative AI tool.  As the magistrate in Warner noted, Rule 26(b)(3)(A) explicitly protects against the disclosure of documents or things “that are prepared in anticipation of litigation or for trial by or for another party.”  A party seeking discovery of this material may still obtain it, but only by showing that it has a “substantial need” for such material and “cannot, without undue hardship, obtain their substantial equivalent by other means.”  Note that while subsection (A) provides a condition allowing disclosure, subsection (B), which applies to mental impressions of a party’s attorneys and representatives, does not.

Because the law on the protection of materials generated by or submitted to consumer-grade generative AI tools remains unsettled, clients should carefully consider whether and how to use such tools in the invention, patent prosecution, and patent litigation contexts.

**************************************************************************************************************

DISCLAIMER: This article is provided by Seed Intellectual Property Law Group LLP (“Seed IP”) for informational purposes only and does not constitute legal advice. The content reflects the views of the authors and not necessarily those of the firm, its partners, or clients. The information provided is “AS IS,” may not reflect the most current legal developments, and is subject to change without notice. We make no representations or warranties regarding accuracy, completeness, or suitability for any purpose. To the fullest extent permitted by law, we disclaim all liability for any loss or damage arising from reliance on this content or linked resources. Accessing or using this article does not create an attorney-client relationship. This content may be considered Attorney Advertising and is subject to our Legal & Privacy notices.

**************************************************************************************************************

For a PDF version of this article, please click here: “Federal Courts Differ on AI Use and Waiver of Work-Product Protection" (coming soon)

About the authors:‍

Jeffrey E. Danley | Partner, Seed IP Law Group

Jessica S. Gritton | Of Counsel, Seed IP Group

John A. Morgan | Partner, Seed IP Law Group

Please feel free to contact us here should you have any comments or questions pertaining to this topic.

‍