When AI Meets Privilege: Early Court Decisions

Two recent decisions have addressed whether materials generated or influenced by AI tools are protected by the attorney-client privilege or work product doctrine. These rulings clarify how privilege claims should be managed when AI is involved, providing key insights for litigants, in-house counsel, and businesses adopting AI in legal workflows.

KEY TAKEAWAYS

• Recent federal court decisions clarify the treatment of the attorney-client privilege and work product doctrine as applied to materials generated using consumer-level (non-enterprise) AI tools.
• Early decisions suggest courts will find that using consumer AI tools may waive the attorney-client privilege, while enterprise AI tools used at the direction of counsel offer more protection.

UNITED STATES V. HEPPNER

Judge Rakoff of the Southern District of New York recently ruled^[1] that seized documents generated by a criminal defendant CEO using a consumer-level (non-enterprise) AI tool after receiving a grand jury subpoena and after retaining counsel were not protected by either the attorney-client privilege or the work product doctrine.

The court reasoned that the attorney-client privilege protects only “confidential” communications between client and counsel (not AI), and that disclosing information to a third-party AI tool—especially a public consumer version that had an express provision that user submissions were not confidential—constituted a disclosure to a third party that destroyed any potential privilege.

The court stated from the bench that the defendant “disclosed it to a third-party, in effect, AI, which had no obligation of confidentiality.” While the court found this lack of confidentiality was fatal to any claim of privilege, the court further noted the absence of privilege given that the defendant’s use of the tool lacked an attorney-client relationship and was not for the purpose of seeking legal advice. 

Similarly, the court found that the “attorney” work product doctrine did not extend to these materials because they did not reflect the legal strategy of counsel nor were they created at the direction of counsel. The court developed this latter factual point through open court questioning of defense counsel regarding their (lack of) contribution or direction before the defendant created the records.

Notably here, given the facts around the defendant’s unilateral use of AI without the involvement of counsel, the court did not address the situation where an attorney directs a client to use an AI tool as part of the legal representation. Further, despite some attempt by defense counsel to emphasize that this work product was the defendant’s own work product and thus protected under Rule 16 of the Federal Rules of Criminal Procedure, the court ruled only on the absence of core “attorney” work product.

In doing so, this Order diverges from other recent non–AI-related work product civil cases[2] that consistently extend work product protection to not only core attorney work product but also, under the plain language of Federal Rule of Civil Procedure 26(b)(3), materials created in anticipation of litigation by a “party or its representative.” The court also did not engage with the related issue of how use of a private AI tool created by lawyers or a legal department to obtain guidance and information might change the analysis.

WARNER V. GILBARCO

By contrast, the Eastern District of Michigan in a civil case^[3] considered the discoverability of materials related to the use of generative AI tools in litigation preparation, drafting a ruling over five pages. The defendants sought production of “all documents and information concerning [the plaintiff’s] use of third-party AI tools in connection with [the] lawsuit.” The court denied this request, finding that such information was work product–protected and therefore not discoverable under Rule 26(b)(3)(A).[4]

The court reasoned that using AI tools to prepare legal materials is analogous to traditional work product–protected activities and rejected the argument that employing generative AI such as ChatGPT amounted to a waiver of work product protection. The court observed that waiver of work product protection requires disclosure to an adversary, or one that makes it likely that materials will reach the hands of an adversary, and that “ChatGPT (and other generative AI programs) are tools, not persons.” The court emphasized that “no cited case orders the production of what Defendants seek here: a litigant’s internal mental impressions reformatted through software.”

The court also denied requests to overrule the plaintiff’s privilege and work-product objections or require a privilege log specific to AI materials, affirming the plaintiff’s right to assert work-product protection and referencing relevant case law.

IMPLICATIONS AND RECOMMENDATIONS

Organizations using AI tools in litigation or internal investigations should be mindful that use of consumer AI platforms may risk waiver of privilege as at least one court has treated public AI tools as third parties for privilege purposes. To minimize risk, organizations should consider using enterprise-grade AI tools with appropriate confidentiality measures and under the direction of counsel (keeping in mind even this does not guarantee protection).

Organizations should review their internal protocols for using AI tools in legal matters, train staff on privilege risks, consider implementing settings that limit or prevent information sharing across the entire organization, and develop documentation practices to support privilege claims. Engaging with counsel to assess privilege strategies and discovery responses can help mitigate risk and ensure compliance with evolving judicial expectations.

HOW WE CAN HELP

The Morgan Lewis eData team stands ready to assist organizations in providing practice frameworks for protection of privilege when rolling out AI tools, including AI risk assessment; internal training on AI use for legal-related tasks; and policy updates.