As litigants and lawyers increasingly experiment with generative artificial intelligence (“AI”) tools, courts are beginning to address an important question: whether a user’s exchanges with an AI platform may be shielded from discovery.
A recent decision from the United States District Court for the Southern District of New York— United States v. Heppner, No. 25 CR. 503 (JSR), 2026 WL 436479 (S.D.N.Y. Feb. 17, 2026)—offers an early example of how courts may approach this issue.
In Heppner, the defendant used a publicly available AI platform to generate reports addressing potential defense strategies and possible responses to anticipated government charges. In connection with a search warrant, the Government seized approximately thirty-one documents memorializing the defendant’s communications with a generative AI platform. Defendant’s counsel asserted that these documents were prepared in anticipation of a potential indictment and were privileged. The Government moved, in writing, for a ruling that the documents were neither protected by attorney-client privilege nor the work-product doctrine. The court ruled that the AI documents do not merit the protections claimed by the defendant.
With respect to the attorney-client privilege, the court held that the communications were not protected because: (a) they were not communications between the defendant and counsel; (b) they were not confidential, particularly in light of the platform’s data-use policies; and (c) they were not made for the purpose of obtaining legal advice from a lawyer. The court noted that the attorney-client privilege requires a “trusting human relationship” with a licensed professional who owes fiduciary duties and is subject to professional discipline.
The court also rejected the defendant’s work-product argument. The court reasoned that the AI-generated materials were not prepared by counsel, at counsel’s direction, nor did they reflect defense counsel’s legal strategy. The Heppner decision applies traditional privilege principles to a new technological context — communications with an AI platform.
However, the court left open the possibility that different facts could produce a different analysis. For example, there may be circumstances in which an AI tool functions more like a professional assistant, consultant, or agent assisting counsel. Whether attorney-client privilege or work-product protection applies in that setting may depend on who is using the tool, for what purpose, under whose direction, and applicable confidentiality protections.
The decision also raises broader questions for litigants and counsel:
- Are AI prompts and outputs meaningfully different from traditional internet searches conducted through Google, Bing, or similar platforms?
- Does the analysis change if the user pays for an enterprise or private AI service that contractually restricts the use of prompts and outputs for training or public disclosure?
- Will courts need to give explicit instructions to litigants to prevent confidential information from being uploaded into AI platforms?
- Should exchanges with AI platforms be treated differently when they are used by parties who are pro se, rather than a party represented by counsel?
In Morgan v. V2X, Inc., No. 25–CV–01991–SKC–MDB, 2026 WL 864223 (D. Colo. Mar. 30, 2026), the United States District Court for the District of Colorado considered whether work product protections apply to a pro se party’s use of AI. Morgan is an employment discrimination case in which the plaintiff was a pro se litigant. The defendant filed a motion to amend the protective order asking the court to (a) specifically address defendant’s AI-related concerns and (b) compel the plaintiff to disclose the identity of the AI tool plaintiff is utilizing in the case. The plaintiff argued that the identity of the AI tool he is using is protected by the work-product doctrine.
The Morgan court concluded that a pro se party has some level of work-product protection in connection with the use of AI platforms. The court reasoned that a reasonable expectation of privacy is not extinguished simply because information is transmitted through an AI platform. The court explained that it is “entirely reasonable” for a person to expect some privacy and confidentiality when interacting with AI tools, even though a third party is collecting and storing their information. Ultimately, the court ruled that the plaintiff must only disclose the name of any AI platform on which he uploaded confidential information. Notably, the court distinguished this case from United States v. Heppner.
As courts continue to confront these questions, it is clear that parties should not assume that communications with generative AI platforms are privileged or protected from discovery merely because they concern legal issues or litigation strategy. Litigants should carefully consider how they are using AI platforms and what information is being entered into the platform.
The discoverability of a party’s exchanges with a generative AI platform is a litigation issue that has not been thoroughly addressed by courts, especially Texas courts. We cannot be certain how Texas courts will address and rule on this issue. However, a Texas court would likely consider ordinary discovery rules and concerns such as relevance, proportionality, possession/control, and privilege.