ARTICLE
18 March 2026

The Risk Of Using Generative AI: Did You Waive Your Attorney-Client Privilege?

FF
Farrell Fritz, P.C.

Contributor

Farrell Fritz is a full-service regional law firm with approximately 80 attorneys in five offices, dedicated to serving closely-held/privately-owned/family owned businesses, high net worth individuals and families, and nonprofit organizations. Farrell Fritz handles legal matters in the areas of bankruptcy and restructuring; business divorce; commercial litigation; construction; corporate and finance; emerging companies and venture capital; employment law; environmental law; estate litigation; healthcare; land use and zoning; New York State Regulatory and Government Relations; not-for-profit law; real estate; tax planning and controversy; tax certiorari, and trusts and estates.

On February 10, 2026, Judge Rakoff of the United States District Court for the Southern District of New York answered that question with a "yes" in United States v Heppner.
United States Technology
Matthew D. Donovan’s articles from Farrell Fritz, P.C. are most popular:
  • in India
Farrell Fritz, P.C. are most popular:
  • within Intellectual Property topic(s)
  • with readers working within the Accounting & Consultancy industries

On February 10, 2026, Judge Rakoff of the United States District Court for the Southern District of New York answered that question with a “yes” in United States v Heppner.

Background

On November 5, 2025, Defendant Benjamin Heppner was arrested on securities and wire-fraud charges. During the investigation, the FBI seized documents, including 31 documents reflecting communications with the AI platform Claude.

The defendant claimed these documents were protected by attorney-client privilege, raising the question of first impression: “whether, when a user communicates with a publicly available AI platform in connection with a pending criminal investigation, are the AI user’s communications protected by attorney-client privilege or the work product doctrine?”

Attorney-Client Privilege

The court found that the communications were not  privileged.

First, the court found that these communications were not communications between a client and a lawyer. The court reiterated that recognized privileges require “a trusting human relationship” and that no such relationship existed, or could exist, between an AI user and an AI platform.

Second,  the court found that the communications were not confidential. In doing so, the court pointed to Claude’s written privacy policy, which explicitly states that the platform reserves its right to disclose data to third parties, defeating any reasonable expectation of confidentiality.

Finally, the court found that the defendant used the AI platform on his own initiative, not at the direction of counsel for purposes of obtaining advice.

Work Product Doctrine

The court also rejected work-product protection. The court found that, even assuming the AI documents were prepared in anticipation of litigation, they were not created at the direction of counsel and could not constitute work product.

Upshot

The decision highlights the privilege risks associated with using generative AI in connection with legal matters. Communications with public AI tools may not be confidential and may be discoverable. One open question is whether the court’s analysis would change if defense counsel had instructed Heppner to obtain information from Claude. But even then, a court might still find that disclosure of attorney-related information to a third-party AI platform waives privilege, particularly when the platform’s terms of service permit data disclosure, but that remains to be seen.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.

[View Source]

Mondaq uses cookies on this website. By using our website you agree to our use of cookies as set out in our Privacy Policy.

Learn More