United States v. Heppner
Summary: Heppner, a businessman in Texas, received a grand jury subpoena and subsequently learned he was the target of the grand jury’s investigation. Heppner, on his own initiative (i.e., without suggestion to do so from his lawyer), then logged on to the “Claude” software platform – a “generative” artificial intelligence chat service from private company Anthropic, to converse with Claude about his situation. Claude prepared for Heppner reports that outlined his defense strategy and what he could argue regarding the facts and law that the government might use against him.
When Heppner was later indicted by the grand jury and arrested for alleged fraud involving investors, the FBI searched his home and seized 31 documents reflecting the communications he had carried on with Claude. Heppner’s lawyer argued these documents were protected from disclosure to the government by both the attorney-client privilege and the work product doctrine. The district court, however, agreed with prosecutors that Heppner’s communications with Claude were neither attorney-client communications nor work product.
Taking up the attorney-client privilege argument first, the court emphasized that the privilege requires an “attorney-client relationship” -- in other words (citing law review commentary), a “trusting human relationship” with a licensed professional who owes fiduciary duties and is subject to discipline. An AI user and platform cannot have such a relationship.
The court also noted that Claude’s terms of use allow collection of user inputs and Claude’s “outputs” and that Anthropic reserves the right to disclose personal data to third parties – even without being compelled by subpoena – in claims and litigation. The court favorably cited another S.D.N.Y. jurist’s conclusion that AI users cannot have substantial privacy interests in conversations with AI that they voluntarily disclose to the platform and that the platform retains in normal course of its business.
Finally, the court rejected the idea that Heppner was communicating with Claude to get legal advice or to talk to counsel, noting that in fact the defendant did not use Claude at his attorney’s suggestion or direction. Moreover, the government noted that Claude, when asked, disclaims being a lawyer or providing legal advice and directs users to consult a “qualified attorney.”
Taking up Heppner’s work product claim, the district court noted that this doctrine protects the mental processes (and material reflecting those processes) of attorneys as they prepare for anticipated or pending litigation. Unfortunately for Heppner, his attorney never asked him to go onto AI and prepare these AI-generated documents. Thus Heppner’s foray into Claude’s world was done on his own volition and could not be considered his attorney’s protected “work product.” The court closed its 12-page opinion by noting that AI’s novelty “does not mean that its use is not subject to longstanding legal principles.”
Opinion: View opinion (PDF, opens in new tab)