When clients use AI: SDNY Court signals boundaries of privilege and work product protection

The written ruling, which followed a bench ruling on February 10, 2026, holds that documents memorializing communications a client had with a generative AI platform and that he later shared with counsel were not protected by the attorney-client privilege or the work product doctrine. The ruling makes clear that communications with AI tools in connection with legal advice and strategy are not automatically protected by the attorney-client privilege or the work product doctrine, particularly where the communications and the records that memorialize them do not reflect communications with counsel, the AI user has no reasonable expectation of confidentiality in communications with the third-party AI platform, and the records were not prepared at the direction of counsel or for the purpose of seeking legal advice.

Background

In United States v. Heppner, the defendant, Bradley Heppner, the founder of an alternative finance firm, was indicted on charges including securities fraud, wire fraud, conspiracy, making false statements to auditors, and falsification of records.

In connection with Heppner’s arrest, agents with the Federal Bureau of Investigation executed a search warrant on Heppner’s home and seized numerous documents and electronic devices. Defense counsel later disclosed to the government that among the seized materials were 31 documents reflecting Heppner’s queries to Anthropic’s AI assistant, Claude, to conduct research related to the government’s investigation and the AI tool’s responses. Without any suggestion from counsel that he do so, Heppner had used Claude to prepare reports outlining defense strategy and arguments in anticipation of a potential indictment. Heppner later provided those materials to his counsel. After he was indicted and the records were seized in connection with his arrest, Heppner, through counsel, asserted the documents were protected by the attorney-client privilege and the work product doctrine. Judge Rakoff rejected both arguments.

The Court’s Decision

Judge Rakoff held that the AI generated documents were not privileged or protected by the work product doctrine because they were not communications between Heppner and his counsel, they were prepared by Heppner in collaboration with a third-party AI platform that did not maintain confidentiality over the communications, and Heppner did not communicate with Claude for the purpose of obtaining legal advice from counsel.

Attorney-Client Privilege

The court emphasized that the attorney-client privilege protects confidential communications between a client and a lawyer (or the lawyer’s agents) conducted for the purpose of obtaining legal advice. Because the AI platform, Claude, was not a lawyer and owed Heppner no fiduciary duty, the communications did not satisfy this threshold requirement.

The court further found that Heppner could have had no reasonable expectation of privacy in his communications with Claude, because he used a consumer AI product whose written privacy policy permitted the provider to use and disclose user data, including communications with the platform, to third parties and governmental authorities.

Work Product Doctrine

The court likewise rejected the argument that the documents were protected by the work product doctrine. Although Heppner argued that the documents were prepared in anticipation of litigation, Judge Rakoff concluded that they did not merit protection under the work product doctrine, because they were not prepared by or at the direction of counsel and did not reflect defense counsel’s strategy at the time that Heppner created them. Sending the AI generated materials to counsel after the fact did not retroactively confer work product protection.

Key Takeaways

While Heppner arises in the context of a criminal case, its reasoning is grounded in core principles of attorney-client privilege and work product protection (i.e., confidentiality, the role of counsel, and reasonable expectations of privacy) that apply equally in civil litigation. As a result, the ruling raises important considerations for both clients and counsel regarding the use of public or consumer AI tools in connection with all legal matters.

In practical terms, the decision highlights the risks associated with clients independently using public generative AI platforms to analyze facts and law or develop litigation strategy, particularly where the platform’s terms allow the provider to access or disclose user inputs.

Notably, although the court found the attorney-client privilege did not protect the AI communications and AI-generated records in this case, the ruling leaves potential pathways for the privilege to apply to AI generated materials under certain circumstances:

• Use of secure, closed AI systems: The court emphasized the public platform’s written policy that it collected data on user inputs and platform outputs, that it used such data to train the model, and that it reserved the right to disclose such data to third parties including governmental and regulatory authorities. This suggests that closed AI environments, where data remains within a secure, isolated boundary and is not used to train the model, may not weaken privilege claims.
• AI document creation at counsel’s direction: The decision also indicates that confidential AI generated materials created at the instruction of counsel could be treated like other client created documents prepared for legal advice or litigation and may therefore qualify for attorney-client privilege or work product protection.

Courts will continue to develop a framework for handling AI generated materials in the months and years ahead. As judges grapple with rapidly evolving AI technologies and long standing legal doctrines, it will be important to monitor how they define the boundaries of attorney-client privilege and work product protection in this new context. Relatedly, as more companies and their executives turn to AI tools for assistance in workstreams, it is important to recognize that communicating with third-party AI tools regarding legal advice or litigation strategy can implicate attorney-client privilege protection. Outside counsel, too, should consider risks inherent in using third-party AI platforms, including as they relate to confidentiality, the attorney-client privilege, and the work product doctrine.
__________

If you have any questions about this Legal Briefing, please feel free to contact any of the attorneys listed or the Eversheds Sutherland attorney with whom you regularly work.