Last week, Liz blogged about a new resource, our “AI Use by Lawyers” Handbook. I figured that, as we were finalizing the Handbook, some development would happen that we’d have to update for before finalizing, but I would have bet that it was some new genAI tool. Instead, it was a court case.

ICYMI, in a February 2026 bench ruling, the U.S. District Court for the Southern District of New York determined that documents outlining a defense strategy and possible legal arguments created by a client using genAI (on his own and not directed by counsel) and sent to his lawyer for review were not protected by privilege or work product doctrine. As this Husch Blackwell alert notes:

This isn’t just a litigation issue. It also applies to preparing for or responding to regulatory audits or other investigations. Feeding legal analysis, or correspondence with counsel or an expert, into an open AI system, potentially waives the attorney-client privilege, confidentiality, and trade secret protections.

This doesn’t necessarily mean any AI use waives privilege. Notably, the defendant’s chats were “with a publicly available, non-enterprise, generative AI platform,” and “not made at the request of counsel.” There’s also the fact that another federal court came out differently on this issue on the same day. As this Proskauer alert notes, in Warner v. Gilbarco, Inc., the Eastern District of Michigan denied a motion to compel discovery of “work product” after a pro se plaintiff used AI tools. What was different (besides the court)? The facts, for one. One case involved the defendant’s use of AI in a criminal trial; the other involved a civil litigant acting as her own counsel. But the courts did seem to apply the same analysis differently:

The [Eastern District of Michigan explained] that work-product waiver requires disclosure “to an adversary or in a way likely to get in an adversary’s hand.” Significantly, the court reasoned that “ChatGPT (and other generative AI programs) are tools, not persons, even if they may have administrators somewhere in the background.” This stands in notable contrast to Heppner, where Judge Rakoff treated the AI platform as a third party for privilege purposes based on its terms of service permitting data disclosure.

The alert concludes:

The Heppner and Warner decisions . . . demonstrate that courts are actively grappling with how traditional privilege and work-product doctrines apply to AI-generated materials. While Heppner reinforces the importance of using properly secured AI tools with confidential or privileged information and ensuring that AI use is directed by counsel, Warner suggests that not all AI-assisted litigation work will ultimately be subject to discovery.

Indeed, despite the seemingly opposite outcomes, both decisions appear to rely on the same basic analytical framework. The critical factors appear to be the specific contractual and technical circumstances of the specific AI platform at issue, whether counsel is involved and/or directed the AI use, whether confidential or privileged information will be entered into the tool, and whether the materials reflect litigation strategy prepared in anticipation of litigation.

On the AI Counsel Blog, Zach shared that:

Ultimately, preserving privilege may boil down to the finer details. Ensuring that AI tools operate in a closed environment, don’t retain data for training, and don’t share data with third parties is a good starting point for creating an expectation of confidentiality. However, until more courts issue rulings on the matter, AI use is risky. Lawyers would be wise to take caution when using AI in conjunction with any sensitive or privileged information.

The Husch Blackwell alert suggests actionable steps that law firms and legal departments should take in light of the decision:

– Update Your Intake: Ask clients whether they’ve discussed their legal matter with any AI tool. Don’t assume they haven’t. Revisit this warning as the matter proceeds.
– Discovery and Depositions: Add AI usage to your deposition questions and collect your client’s AI chats. Listing the chats on a privilege log triggered the issue in Heppner. Expect such scrutiny going forward.
– Educate Clients: Make it clear that using public AI tools to summarize a legal memo from counsel— or to process or brainstorm legal matters—can waive privilege, confidentiality, and trade secret protections.
– Train Your Teams: Regulatory, compliance, and operations staff should be warned that running confidential or trade secret info through public AI is a privilege and confidentiality risk is waiting to happen.
– Stick to Enterprise Tools: Company-approved, closed-universe AI tools are different, but review their terms and train users accordingly.

Be sure to subscribe to our AI Counsel Blog, where John and Zach have discussed these topics a few times and generally roll up their sleeves to address some of the more granular issues that legal and compliance personnel are confronting when trying to manage the risks of emerging technologies.

– Meredith Ervine