A Federal Court on AI and Privilege

On February 17, 2026, Judge Jed Rakoff of the Southern District of New York issued what appears to be the first federal decision directly addressing whether communications with a generative AI platform are protected by attorney-client privilege or the work product doctrine. His answer was straightforward: they are not. 

The case, United States v. Heppner, arose in a criminal context. A defendant, on his own initiative, had used Anthropic’s Claude to generate written analyses relating to his defense strategy. After his electronic devices were seized pursuant to a search warrant, the government sought access to those AI interactions. The defendant argued that the communications with "Claude" were protected either by attorney-client privilege or by the work product doctrine. The Court rejected both arguments.

Click to read the 12-page Memorandum by Judge Rakoff

Importantly, the ruling did not hinge on hostility toward AI. It turned on traditional doctrine. Attorney-client privilege protects confidential communications between a client and a lawyer for the purpose of obtaining legal advice. Claude, the Court noted, is not a lawyer, and no attorney-client relationship existed between the defendant and the platform. The fact that the defendant may have intended to use the output in later discussions with counsel did not transform the original interaction into a privileged communication.

The Court also examined confidentiality. Because the platform’s terms permitted data collection and potential disclosure, the defendant could not reasonably claim that the interaction carried the level of confidentiality that privilege doctrine requires. Sharing the AI-generated materials with counsel after the fact did not retroactively cloak them in protection.

The work product argument fared no better. Work product doctrine is designed to protect the mental processes of attorneys and materials prepared at their direction in anticipation of litigation. Here, the documents were generated independently by the defendant, not at the direction of counsel. The Court emphasized that longstanding limits on the doctrine do not dissolve simply because a new technological intermediary is involved.

The broader signal is not that AI cannot be used in legal practice. It is that courts will evaluate AI interactions using established legal frameworks. Novelty does not suspend doctrine. A tool may be sophisticated, but the questions remain familiar: Who was involved? What was the purpose? Was it confidential? Was counsel directing the process?

For firm leadership, the lesson is not technological. It is structural. When lawyers, staff, or clients use public AI tools in connection with active matters, the legal analysis will not turn on marketing claims about security or innovation. It will turn on whether the traditional elements of privilege and work product are satisfied under the facts.

New tools may accelerate workflows. They do not relax professional standards. Courts will apply existing law to contemporary behavior.

Leadership’s role is to ensure that behavior has been thought through before a court is asked to examine it.

Rex C. Anderson
Desert Sage AI
AI Governance for Law Firms and Accounting Firms

__________________________________

Forwarded to you? 
Subscribe for direct delivery of future issues.

 Subscribe