When a client uses a generative AI platform to prepare materials for discussions with counsel in a litigation preparation context, are those materials protected by the attorney-client privilege or the work product doctrine? In United States v. Heppner, Judge Jed S. Rakoff of the Southern District of New York answered with a categorical no. The ruling appears to be the first of its kind in federal court — and its reasoning and categorical rule have implications far beyond the facts of the case.

In a new Law360 Expert Analysis, Matt Coogan and Cindy Kuang argue that Heppner‘s broad holding rests on an unstable premise — that an AI platform is a “third party” — and that the opinion misreads the very Anthropic consumer policies on which it relies, including by characterizing the privacy policy as authorizing training on user data when, in fact, the policy expressly prohibits it.

The article identifies significant gaps in the court’s work-product and confidentiality analyses and makes the affirmative case that existing doctrine, including the Second Circuit’s frameworks in DeFonte and Kovel and the SDNY’s own Asia Global Crossing test, already provides the tools courts need to resolve these questions without resorting to categorical rules.

The article also situates Heppner against the Eastern District of Michigan’s holding in Warner v. Gilbarco, which found that use of a generative AI platform does not waive work-product protection. This underscores the divergent analytical approaches federal courts are taking on these questions.

The full article is available on Law360 or at the PDF link on this page.