Having clients run AI queries about their case and make suggestions or ask follow up questions is familiar terrain for most attorneys these days. But are those queries and answers privileged? No, says prominent federal judge Jed Rakoff of the Southern District of New York.
In United States v. Hepner, the FBI searched the home of the defendant, who was charged with securities and wire fraud. It seized a number of electronic devices and found 31 documents memorializing communications between the defendant and Claude AI, in which the defendant outlined defense strategy and arguments he could make in response to what the government might charge him. The defendant’s lawyers argued that the communications were privileged. They were prepared in anticipation of criminal charges. They reflected things his attorneys had told him. They were created for the purpose of sharing with his attorneys and were shared with his attorneys.
Nevertheless, the court held the communications were not privileged under either the attorney-client privilege or the work product doctrine. The court held that Claude is not an attorney and, therefore, communications with it do not qualify for attorney-client privilege protection. Neither, concluded the court, were the communications confidential. Claude gathers data from such inputs to train its system and notifies users that it may disclose personal data to third parties. The court also rejected the contention that the communications were for the purpose of obtaining legal advice. It noted that the outcome might have been different if the defendant’s counsel had directed him to use Claude. But communications do not become privileged because they are subsequently shared with a lawyer.
The same reasoning torpedoed the defendant’s work product doctrine argument. The work product doctrine protects from disclosure materials prepared “in anticipation of litigation,” but only “at the behest of counsel.” Because the communications were not made at the behest of counsel, they did not qualify for protection. The defendant acted “on his own” and, therefore, could not claim either attorney-client or work product protection.
These interactions are commonplace today and counsel should be cautioning their clients at the outset of an engagement about the risk that their AI interactions may be discoverable.
Sean Carnathan
Are Your Client’s AI Queries Privileged?
Having clients run AI queries about their case and make suggestions or ask follow up questions is familiar terrain for most attorneys these days. But are those queries and answers privileged? No, says prominent federal judge Jed Rakoff of the Southern District of New York.
In United States v. Hepner, the FBI searched the home of the defendant, who was charged with securities and wire fraud. It seized a number of electronic devices and found 31 documents memorializing communications between the defendant and Claude AI, in which the defendant outlined defense strategy and arguments he could make in response to what the government might charge him. The defendant’s lawyers argued that the communications were privileged. They were prepared in anticipation of criminal charges. They reflected things his attorneys had told him. They were created for the purpose of sharing with his attorneys and were shared with his attorneys.
Nevertheless, the court held the communications were not privileged under either the attorney-client privilege or the work product doctrine. The court held that Claude is not an attorney and, therefore, communications with it do not qualify for attorney-client privilege protection. Neither, concluded the court, were the communications confidential. Claude gathers data from such inputs to train its system and notifies users that it may disclose personal data to third parties. The court also rejected the contention that the communications were for the purpose of obtaining legal advice. It noted that the outcome might have been different if the defendant’s counsel had directed him to use Claude. But communications do not become privileged because they are subsequently shared with a lawyer.
The same reasoning torpedoed the defendant’s work product doctrine argument. The work product doctrine protects from disclosure materials prepared “in anticipation of litigation,” but only “at the behest of counsel.” Because the communications were not made at the behest of counsel, they did not qualify for protection. The defendant acted “on his own” and, therefore, could not claim either attorney-client or work product protection.
These interactions are commonplace today and counsel should be cautioning their clients at the outset of an engagement about the risk that their AI interactions may be discoverable.