In an opinion issued February 17, 2026, Judge Jed Rakoff of the Southern District of New York ruled that communications between a criminal defendant and a generative AI platform not directed by counsel are not protected by the attorney-client privilege or the work product doctrine. The ruling in United States v. Heppner appears to be the first of its kind nationwide and carries significant implications for how individuals utilize generative AI when they become aware they are under investigation. Bienert Katzman Littrell Williams LLP attorney C.B. Rome analyzes the opinion and its potential impact.

The Facts

Bradley Heppner, a former executive charged with securities fraud, wire fraud, and related offenses, had approximately thirty-one documents seized from his home that memorialized his conversations with Claude, an AI platform operated by Anthropic. Critically, these communications occurred after Heppner had received a grand jury subpoena and learned he was the target of a federal investigation, but before he was indicted. Heppner argued the documents were privileged because he created them in anticipation of speaking with his attorneys. The court disagreed.

Key Takeaways

AI platforms are not confidential. Judge Rakoff emphasized that users of AI platforms like Claude consent to privacy policies that allow the provider to collect user inputs and outputs, use that data for training, and disclose it to third parties, including government authorities, even without a subpoena. Under these circumstances, no reasonable expectation of confidentiality exists.

The purpose matters, and so does counsel’s direction. The court noted that Heppner communicated with Claude on his own initiative, not at counsel’s direction. Had his attorneys instructed him to use the AI platform, Judge Rakoff suggested it “might arguably” have functioned like a professional agent acting within the protection of the privilege. Because Heppner acted independently, the relevant inquiry was whether he sought legal advice from Claude, and Claude explicitly disclaims providing such advice.

Work product protection requires counsel’s involvement. The work product doctrine shields materials prepared “by or at the behest of counsel” that reflect counsel’s strategy. Because Heppner created the AI documents on his own volition, without direction from his attorneys, and the documents did not reflect his counsel’s strategy at the time they were created, the doctrine offered no protection.

Practical Guidance

For individuals who learn they are under investigation, Heppner delivers a clear message: do not use AI platforms to analyze your legal situation or prepare defense materials unless your attorney specifically directs you to do so, and even then, exercise caution. What you share with an AI is not confidential, and those communications may be seized and used against you.

The temptation to turn to AI for quick answers is understandable—but when your liberty and livelihood are at stake, there is no substitute for experienced legal counsel who can protect your rights, maintain true confidentiality, and develop a defense strategy tailored to your circumstances. AI cannot do that. The earlier you engage knowledgeable counsel, the better positioned you will be to protect yourself.