On February 10, 2026, Magistrate Judge Anthony P. Patti of the U.S. District Court of the Eastern District for Michigan ruled in Warner v. Gilbarco (“Warner”) that work product protection applied to materials generated by a third-party artificial intelligence (“AI”) tool prepared by a pro se plaintiff in the course of litigation.[1] In denying the defendants’ requests for discovery related to plaintiff’s use of third-party AI tools, Judge Patti warned against the over fixation on a party’s mere use of AI tools, explaining that “[generative AI programs] are tools, not persons” and noting that if the simple act of uploading information onto an AI platform waived protection, then this “would nullify work-product protection in nearly every modern drafting environment, a result no court has endorsed.”[2] Judge Patti’s decision came on the same day as the initial bench ruling in the criminal matter U.S. v. Heppner (“Heppner”), in which Judge Rakoff of the U.S. District Court for the Southern District of New York held that documents a defendant had generated using a third-party AI tool were not protected by either the attorney-client privilege or the work product doctrine.[3]
Key Takeaways
- Courts may characterize AI differently. Whether an AI tool is characterized as a “tool” or as a third-party capable of receiving disclosures may have a significant impact on any ruling. In Warner, the court’s conclusion that AI programs “are tools, not persons” supported a finding that work product protection was not waived. In Heppner, the court treated disclosure to AI as a third-party disclosure that defeated expectations of confidentiality.
- Counsel’s direction may impact the outcome. When represented by counsel, whether a party utilized an AI tool at the direction of counsel may be a critical factor in determining work product protection. In Heppner, Judge Rakoff noted that the work product doctrine was intended to “shelter the mental process of the attorney;” the doctrine was not intended to “shield[] from discovery material in an attorney’s possession that were prepared by neither the attorney nor his agents.”
- Differing interpretations of the scope of the work product doctrine may matter. The court’s interpretation of the scope of the work product doctrine, including whether the doctrine is primarily about a “party” generating material or about attorneys’ (and their agents’) mental impressions may be an important consideration in determining what is protected under the work product doctrine. In Warner, Judge Patti gave significance to the word “party” in Federal Rule of Civil Procedure 26(b)(3)(A) and acknowledged that a pro se litigant qualified as such. In Heppner, a criminal matter where the civil rules do not apply, Judge Rakoff stated that the work product doctrine “shelters the mental process of the attorney” or his agents.
- Analysis varies based on the nature of the protection claim. Waiver analysis will differ significantly for assertions of the attorney-client privilege and work product protection. As Judge Patti explained, “while the mere showing of a voluntary disclosure to a third person will generally suffice to show waiver of the attorney-client privilege, it should not suffice in itself for waiver of the work product privilege.” Accordingly, the court was unconcerned with the fact that the AI tool “may have administrators somewhere in the background.”
- Privacy policies may or may not factor into the analysis. The relevance of an AI platform’s terms of service remains unsettled. In Heppner, the court relied on the platform’s privacy policy in finding that the defendant had no reasonable expectation of confidentiality. In Warner, while the court did not analyze any privacy policy or terms of service, it seemed unconcerned with the use of such a tool from a waiver perspective. This divergence raises questions about whether standard click-through terms should be dispositive of confidentiality.
Background
As a general matter, the attorney-client privilege protects “[c]ommunications from attorneys to their clients that contain legal opinions and legal recommendations, and are intended to be confidential,” including “communications that are made for the purpose of obtaining or providing legal advice.”[4] The burden of establishing the existence of the privilege rests with the party asserting it.[5]
The work product doctrine is routinely articulated as providing protection for materials that are prepared by counsel, or at counsel’s direction, if they were prepared because of a party’s objectively reasonable subjective anticipation of litigation.[6] The Federal Rules of Civil Procedure, however, explain that the doctrine protects “documents and tangible things that are prepared in anticipation of litigation or for trial by or for another party or its representative,” such as an attorney.[7]
The Warner Decision (E.D. Mich.)—Work Product Protection Applies
On September 6, 2024, a pro se plaintiff brought an employment discrimination lawsuit against her former employer, Gilbarco, Inc., and its parent company, Vontier Corporation. On December 23, 2025, the defendants filed a motion to compel, which included (1) a request seeking production of “all documents and information concerning [Plaintiff’s] use of third-party AI tools in connection with this lawsuit” and (2) a request to the court to “‘[o]verrule[ ] Plaintiff’s attorney–client privilege and work-product objections to the AI materials,’ or alternatively, ‘if the Court sustains any privilege or protection as to any item, require Plaintiff, within seven days, to serve a Rule 26(b)(5)(A)-compliant privilege log.’”[8]
Work Product Protection Applied to Pro Se Litigant’s AI-Generated Materials
Judge Patti’s analysis was grounded in Rule 26(b)(3)(A) of the Federal Rules of Civil Procedure, which provides that “[o]rdinarily, a party may not discover documents and tangible things that are prepared in anticipation of litigation or for trial by another party or its representative[.]” Judge Patti asserted that information concerning the plaintiff’s use of third-party AI tools in connection with the lawsuit was not discoverable, qualifying the plaintiff’s chats with the AI tool as “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.” The court relied on precedent recognizing that pro se litigants are entitled to work product protection for their mental impressions, drafts, and internal litigation strategy. Notably, outside attorney involvement was not treated as a prerequisite to work product protection.[9]
AI as a “Tool”—Disclosure to AI Does Not Waive Work Product Protection
Judge Patti made clear that work product protection is not automatically waived by a disclosure. Critically, he clarified that “the work-product waiver has to be a waiver to an adversary or in a way likely to get in an adversary’s hand.” Because generative AI programs “are tools, not persons, even if they may have administrators somewhere in the background,” the court concluded that use of an AI tool did not constitute the kind of third-party disclosure that would defeat protection. This tool-based characterization was important to the outcome: treating AI as a tool rather than a person meant that the plaintiff’s use of the AI tool did not waive work product protection.[10]
Judge Patti accordingly upheld the plaintiff’s assertion that work product protection applied to information concerning her use of AI tools in preparation of litigation, agreeing with the plaintiff that the “Defendants’ request is a fishing expedition.” He also noted that the defendants’ theory that work product protection was waived by the plaintiff’s use of an AI tool “would nullify work-product protection in nearly every modern drafting environment, a result no court has endorsed.”[11] Notably, Judge Patti did not analyze or rely on any AI privacy policy or terms of service in reaching his decision. The absence of such an analysis means that confidentiality concerns tied to platform terms did not factor into discoverability determination under Rule 26(b)(3)(A).
Discovery of Pro Se Plaintiff’s Use of AI Tools Is Neither Relevant nor Proportional
Additionally, Judge Patti explained that, under Fed. R. Civ. P. 26(b)(1), the defendants’ request for information concerning the plaintiff’s use of AI tools “is not relevant, or, even if marginally relevant, is not proportional.” He noted that the court had “previously confirmed that Defendants have no evidence of Plaintiff having violated the protective order by uploading documents marked confidential onto an AI platform.” Judge Patti also commented that in his review of the plaintiff’s deposition transcript, he found “an inordinate amount of questioning about Plaintiff’s use of AI, but again, no suggestion that she uploaded prohibited items” and warned the defendants that their “preoccupation with Plaintiff’s use of AI needs to abate.”[12]
The Heppner Decision (S.D.N.Y.)—No Privilege or Work Product Protection
On February 10, 2026, Judge Jed S. Rakoff of the U.S. District Court of the Southern District of New York ruled in U.S. v. Heppner that documents a defendant had generated using a third-party AI tool were not protected by either the attorney-client privilege or the work product doctrine. In that case, the defendant had used an AI platform, without any suggestion or direction from counsel, to prepare reports that outlined defense strategy in anticipation of a pending criminal indictment. His counsel subsequently asserted privilege and work product protection over these AI-generated documents.[13] Judge Rakoff concluded that the AI-generated documents were not privileged for three reasons: (1) the AI tool is not an attorney, and “[n]o such relationship exists, or could exist, between an AI user and [the AI platform]”; (2) the communications were not confidential because, under the AI platform’s privacy policy, the platform “reserves the right to disclose” user data to third parties, meaning the defendant could have had no “reasonable expectation of confidentiality in his communication”; and (3) the defendant did not use the AI tool at the direction of counsel, so the communications were not made for the purpose of obtaining legal advice. Judge Rakoff further ruled that the work product doctrine did not apply because the documents “were prepared by the defendant on his own volition” and were not “prepared by or at the behest of counsel” nor did they “reflect defense counsel’s strategy.”[14]
Comparing Warner and Heppner—Key Differences in Approach
Interpretation of the Work Product Doctrine
The Warner and Heppner decisions reach seemingly conflicting outcomes because of fundamentally different interpretations of the work product doctrine and different characterization of AI. In Warner, Judge Patti adopted a literal, party-centric interpretation of Rule 26(b)(3)(A), giving independent significance to the word “party” and recognizing that a pro se litigant qualifies as a party capable of generating protected work product. However, the court made no suggestion that the plaintiff’s pro se status affected the work product analysis.
In Heppner, on the other hand, Judge Rakoff took the view that the work product doctrine is about attorneys’ mental impressions. In particular, he emphasized that “[a]t its core[,]…[the work product doctrine] shelters the mental process of the attorney” and that the doctrine’s purpose “is not generally promoted by shielding from discovery material in an attorney’s possession that were prepared neither by the attorney nor his agents.” Notably, the work product doctrine applies more narrowly in the criminal context than in civil litigation. As a criminal matter, the Federal Rules of Civil Procedure—including Rule 26(b)(3)(A), which extends work product protection to materials prepared “by or for another party of its representative”—did not apply.[15] Judge Rakoff therefore did not have reason to weigh the broader language of Rule 26(b)(3)(A) in making his ruling, and instead applied the more attorney-centric formulation of the doctrine as standard in criminal proceedings.
Characterization of AI: “Tool” vs. Third-Party
The fundamental analyses presented in Warner and Heppner demonstrate two contrasting approaches with regards to privilege and work product doctrine analysis. In Warner, Judge Patti categorized AI at the outset as a “tool,” stating that “[generative AI programs] are tools, not persons.” Because waiver generally requires disclosure to an adversary, or disclosure in a manner likely to reach an adversary, treating AI as a tool meant that its use did not constitute a third-party disclosure that would undermine protection. By contrast, in Heppner, Judge Rakoff did not explicitly engage with this threshold question but instead emphasized that “[b]ecause [the AI tool] is not an attorney, that alone disposes of Heppner’s claim on privilege.” While Judge Rakoff’s finding went to the court’s holding on assertions of the attorney-client privilege and not work product, the differing views were important to the courts’ different rulings and may play a significant role in future disputes exploring similar issues.
The Role of Privacy Policies and Terms of Service
Another difference in Warner and Heppner is that while Judge Rakoff took into consideration the privacy policies and terms of service of the relevant AI platform, Judge Patti did not. Specifically, in Warner, Judge Patti applied the traditional rule that work product is waived only by disclosure to an adversary or in a manner likely to reach an adversary and did not analyze or rely on any AI privacy policy or terms of service. In Heppner, by contrast, Judge Rakoff treated the defendant’s use of a publicly available AI platform—particularly in light of its terms of service—as defeating confidentiality and supporting discoverability. Judge Rakoff noted that the AI platform’s privacy policy “reserves the right to disclose” user data to third parties, including “governmental regulatory authorities,” and concluded that the defendant therefore had no “reasonable expectation of confidentiality.”
This divergence raises an unresolved question for future courts: whether consent to disclosure embedded in boilerplate language in terms of service or privacy policies—some routinely accepted through click-through agreements, others merely posted on a service provider’s website—should meaningfully determine confidentiality for work product or privilege purposes. Raising the question of whether ubiquitous provisions found in modern cloud-based technology agreements, standing alone, demonstrate a litigant’s affirmative intent to disclose litigation strategy.
Conclusion
The Heppner and Warner decisions highlight the unsettled state of the law regarding AI use, attorney-client privilege, and the work product doctrine. While Heppner declined to extend privilege and work product protection to materials a criminal defendant created using a publicly available AI tool without counsel’s direction, Warner upheld work product protection for similar materials for a civil pro se litigant, reasoning that AI tools “are tools, not persons.” Different facts—including counsel involvement, the nature of the AI tool, and the specific privilege or protection asserted—will likely lead to different results. Parties and counsel should proceed cautiously, particularly when using consumer-facing AI tools, and should be mindful that courts may find that how an AI tool is used, who uses it, and under what circumstances, may affect determinations of if and how that material is protected.
We will continue to monitor developments related to the impact of using AI tools on litigation and provide further updates as warranted.
* * *
[1] See Warner v. Gilbarco, Inc., No. 2:24-CV-12333, 2026 WL 373043 (E.D. Mich. Feb. 10, 2026).
[2] Id. at *4 (emphasis added).
[3] Transcript of Record at 2:25-3:15, U.S. v. Heppner, No. 25 CR. 503 (JSR) (S.D.N.Y. Feb. 10, 2026); see also United States v. Heppner, No. 25 CR. 503 (JSR), 2026 WL 436479 (S.D.N.Y. Feb. 17, 2026). We previously addressed the Heppner decision in detail. See Paul, Weiss, S.D.N.Y. Court Considers Whether AI-Generated Documents Are Subject to Privilege Protections (Feb. 20, 2026), available here.
[4] AXIS Ins. Co. v. Innovation Ventures, LLC, No. 08-15298, 2010 WL 11541982, at *2 (E.D. Mich. Feb. 8, 2010).
[5] United States v. Dakota, 197 F.3d 821, 825 (6th Cir. 1999).
[6] Hillman Power Co., LLC v. On-Site Equip. Maint., Inc., 672 F. Supp. 3d 379, 386–87 (E.D. Mich. 2023).
[7] Fed. Rule Civ. Proc. 26(b)(3)(A).
[8] Warner, 2026 WL 373043 at *4.
[9] Id.
[10] Id. at *4 (emphasis added).
[11] Id. at *4.
[12] Id. at *4; n3.
[13] Heppner, 2026 WL 436479, at *1.
[14] Id.
[15] United States v. Jimenez, 265 F.Supp.3d 1348 (S.D. Ala. 2017) provides an example of a federal district court explicitly rejecting a defendant’s attempt to invoke Federal Rule of Civil Procedure 26(b)(3)(A) and related civil discovery protections in a criminal case. This is further confirmed by other courts. See e.g., United States v. Thomas, 769 F. Supp. 3d 739 (N.D. Ohio 2025); United States v. Forman, 2021 WL 520045 (E.D. Tex. Feb. 10, 2021).
Related Insights
