Three federal courts have now weighed in on whether privilege and work product protections apply to litigants’ use of generative AI tools. The short answer: it depends on what type of protection you’re claiming, who is claiming it, and what the platform’s terms of service say. Here is what in-house counsel and outside advisors need to know.
Attorney-client privilege over AI communications is a losing argument
In United States v. Heppner (S.D.N.Y. Feb. 17, 2026), a represented criminal defendant used the public version of Anthropic’s Claude to prepare defense strategy documents on his own initiative, then claimed privilege over the resulting exchanges. The court rejected the claim on every element: Claude is not an attorney, Anthropic’s privacy policy permitted data collection and third-party disclosure (destroying any expectation of confidentiality), and Claude itself disclaimed providing legal advice. Assume that any communication with a public AI platform will not be protected by attorney-client privilege. The elements simply cannot be met.
Work product protections are more durable—but not automatic
Both Warner v. Gilbarco, Inc. (E.D. Mich. Feb. 10, 2026) and Morgan v. V2X, Inc. (D. Colo. Mar. 30, 2026) found that materials created using AI can qualify for work product protection under Federal Rule of Civil Procedure 26(b)(3). The Warner court held that AI systems are “tools, not persons,” so disclosure to them is not disclosure to an adversary—the standard trigger for waiver. The Morgan court agreed, extending this reasoning to pro se litigants and adding a notable privacy analysis: citing United States v. Warshak and Carpenter v. United States, the court concluded that “routing information through a third-party system does not forfeit all privacy” and that modern AI platforms, which “invite candid and significant disclosure of information” and “simulate empathy, foster trust, and interact in a way that feels genuine and intimate,” may actually support stronger privacy expectations than traditional internet services.
However, work product protection is not unlimited. The Morgan court required the plaintiff to disclose the identity of the AI tool used, finding that simply naming a platform does not reveal mental impressions or legal strategy. The protection covers the substance of AI interactions—the prompts, outputs, and iterative analysis—but not the fact of AI use itself.
The platform’s terms of service are now a litigation risk factor
Across all three decisions, the AI provider’s terms of service played a material role. In Heppner, Anthropic’s privacy policy—which permits training on user inputs and disclosure to third parties under the defendant’s subscription—was central to the finding that no confidentiality existed. In Morgan, the court’s amended protective order established what amounts to a contractual floor for permissible AI use with confidential information: the provider must be contractually prohibited from using inputs to train its model, from disclosing inputs to third parties except as essential to service delivery, and must afford users the ability to delete all confidential information upon request. Mainstream free-tier AI tools will not meet this standard. Enterprise agreements may.
Practical steps for counsel
Review your AI provider agreements now. Determine whether your organization’s AI tools meet the Morgan standard: no training on inputs, no third-party disclosure, and the right to delete. If they don’t, reconsider what information flows through them.
Distinguish between privilege and work product when advising clients. Attorney-client privilege requires confidentiality, a communication with counsel, and a legal-advice purpose—elements public AI use will almost certainly defeat. Work product protection asks a different question: whether disclosure was to an adversary. These are different analyses with different outcomes, and the advice should reflect that.
Revisit your ethical obligations. Model Rule 1.6(c)’s requirement to make “reasonable efforts” to prevent unauthorized disclosure of client information, and Model Rule 1.1’s duty of technological competence, apply squarely here. Understanding how an AI tool handles data is no longer optional—it is part of the duty of competence.
Anticipate protective order provisions addressing AI. The Morgan court drafted its own AI-specific protective order language when it found both parties’ proposals inadequate. Expect opposing counsel and courts to raise AI use in discovery disputes with increasing frequency. Being prepared with documentation of your platform’s contractual safeguards will be more persuasive than scrambling to respond after the fact.
Monitor the access gap. The Morgan court candidly acknowledged that its standard will “at least for now” bar most mainstream free AI tools from processing confidential information, disadvantaging pro se litigants and smaller organizations. This tension between data security and equitable access will shape the next wave of decisions.
The law on AI and privilege is forming quickly. Heppner, Warner, and Morgan are not the last word, but they are the first—and the framework they establish will matter. The duty of competence now includes knowing not just whether to use AI, but which AI to use and on what terms.
YES! PLEASE SIGN ME UP TO RECEIVE EMAIL ALERTS FROM OTHER GUNSTER PRACTICE AREAS.
Gunster. Florida's Law Firm for Leaders.
As a full-service law firm, Gunster provides legal counsel to leading organizations and individuals from its 13 offices statewide. Established in 1925, the firm has expanded, diversified, and evolved, but always with a singular focus: Florida and its clients’ stake in it. A magnet for business-savvy attorneys who embrace collaboration for the greatest advantage of clients, Gunster’s growth has not been at the expense of personalized service but because of it. The firm serves clients from its offices in Boca Raton, Coral Gables, Fort Lauderdale, Jacksonville, Miami, Naples, Orlando, Palm Beach, Stuart, Tallahassee, Tampa, Vero Beach, and its headquarters in West Palm Beach. With more than 320 attorneys and consultants and 300 committed support staff, Gunster is ranked among the top 200 largest law firms by the National Law Journal and has been recognized as one of the Top 100 Diverse Law Firms by Law360. More information about its practices, industries, offices, and news is available at www.gunster.com.
