News & Media
Publications

The recent explosion of AI use has made legal analysis available to everyone.  Clients and litigants who formerly had to rely on their attorneys for virtually all their legal analysis now may query an AI platform and obtain some fairly accurate information on their own.  This access may undermine the long-established doctrines of attorney-client privilege and the attorney work product doctrine. 

The legal doctrine of attorney-client privilege protects communications between a client and their attorney from being disclosed to third parties or compelled in evidence in legal proceedings.  It is one of the most fundamental privileges in common law.  The protection extends to all forms of communication between an attorney and their client when the client is seeking legal counsel, with certain exceptions: communications made in furtherance of a crime or fraud, information about a fact in communication that exists independently, and business advice as opposed to legal advice. Importantly, sharing communications with unnecessary third parties waives the privilege.  The attorney-client privilege is held by the client so only the client can waive it.  Thus, if the client voluntarily discloses the communication to a third party, the privilege is lost.

Related to the attorney-client privilege is the attorney work product doctrine, a legal protection that shields an attorney’s case preparations and legal strategies from being shared with opposing parties during a lawsuit.  The attorney work product doctrine allows attorneys to build their cases and investigate claims without fearing that their thought processes will be exposed.  For documents or materials to qualify as a work product, they must be created specifically in preparation of a lawsuit or a realistic possibility of impending litigation.

While related, the attorney-client privilege and attorney work product serve different purposes.  The attorney-client privilege protects confidential communications between an attorney and their client, while the attorney work product doctrine protects the materials and documents prepared by the legal team, including items that were never shared with the client. Together, these doctrines ensure attorneys may represent their clients vigorously and free from the prying eyes of opponents. 

In what was described by the presiding judge as “a question of first impression nationwide,”[1] in the United States v. Heppner,  Judge Rakoff of the Southern District of New York ruled that written exchanges between a criminal defendant and the generative AI platform Claude were not protected from plaintiff government’s inspection by either the attorney-client privilege or the attorney work product doctrine. 

Defendant Heppner was indicted on various charges related to securities and corporate fraud.  In connection with his arrest, the FBI executed a search warrant at his house and seized numerous documents and electronic devices, among them many documents that memorialized communications about the case that Hepper had with Claude, the AI platform owned by Anthropic. Without any suggestion from counsel that he do so, Heppner queried Claude about various issues related to his defense strategy.  Heppner, through his counsel, subsequently tried to assert the attorney-client privilege over these documents, arguing that he had inputted into Claude information that he had learned from counsel and that he had created the AI documents for the purpose of speaking with counsel to obtain legal advice.  In response, the government moved for a ruling that the AI documents were protected by neither the attorney-client privilege nor the work product doctrine. 

In crafting its ruling related to attorney-client privilege, the judge noted that “[c]ourts construe the attorney-client privilege narrowly because it operates as an exception to the rule that all relevant proof is essential for a complete record and for confidence in the fair administration of justice.[2]  The judge referenced the well-established three-part attorney-client privilege test that would apply to any communication by a client: (1) between client and attorney, (2) intended to be and were kept confidential, and (3) for the purpose of obtaining or providing legal advice.  In applying these principals the judge found that the AI documents were not communications between Heppner and his counsel because Claude is not an attorney.  Also, the AI documents were not confidential both because Heppner communicated them to a third-party-the AI platform-and importantly, because Anthropic’s consent form provides that the data regarding the queries into Claude and its responses are used by Anthropic to train Claude and that Anthropic reserves the right to disclose such data to third parties, including governmental regulatory authorities.  Thus, the government was allowed to access the communications Heppner had with Claude.

Regarding the attorney work product doctrine, the Heppner court found that the documents did not warrant those protections because they were not prepared by Heppner’s counsel, or at his counsel’s request.  Instead, Heppner prepared those documents on his own, so they did not reflect his counsel’s strategy at the time they were created. 

In contrast, in Warner v. Gilbarco,[3] Inc, a case decided virtually concurrently with Heppner, the Court found that records reflecting a pro se litigants use of ChatGPT to assist with legal filings were protected by the work-product doctrine. In this matter, a former employee brought discrimination and related claims against her employer, Gilbarco, Inc. Notably, the plaintiff represented herself in the litigation and acknowledged during discovery she used ChatGPT to help draft filings. The defendant employer sought broad discovery of “all documents and information” concerning the plaintiff’s use of third-party AI tools in connection with the litigation. The plaintiff objected, arguing that such materials reflected her internal analysis and mental impressions, and were therefore protected from disclosure.  The court rejected the defendant’s arguments, concluding that the plaintiff’s interactions with ChatGPT were shielded by the work-product doctrine. Importantly, the court emphasized that pro se litigants are not categorically excluded from asserting work-product protection.

In considering these cases, it is important to note that the plaintiff in the Warner case was representing herself, effectively she was both the attorney and the client.  In addition, the plaintiffs in both cases apparently were using consumer-level subscriptions to the respective AI platforms. Most AI platforms offer specific tiers with zero data retention, and had those subscriptions been used, the Heppner analysis may have changed.  It remains to be seen how future courts will rule when these issues arise.