United States v. Heppner: The Case That Changed AI Legal Privilege

Short answer: In United States v. Heppner, a February 17, 2026 ruling, Judge Jed Rakoff held that Bradley Heppner's AI-made documents were not protected. Privilege failed because Claude is not a lawyer and, under the platform's privacy policy, the chats were not confidential. Work product protection failed because he acted without attorney direction. The ruling does not bind Arizona courts.

A federal court decision in February 2026 changed how legal professionals think about AI and client confidentiality. The case is United States v. Heppner. Judge Jed Rakoff of the Southern District of New York decided it. It was the first major ruling on a key question. Do talks with AI platforms get the same legal protections as talks between attorneys and clients? The court's answer was no.

The ruling binds only that case. But it is a useful warning for anyone who shares legal matters with an AI tool, including Arizona clients and lawyers.

The Heppner Case: When AI Strategy Backfires

Bradley Heppner found himself in serious legal trouble. He was later charged with securities fraud, wire fraud, and conspiracy related to his role as chairman of GWG Holdings. Before those charges, he received a grand jury subpoena and learned he was the target of the investigation. Without any suggestion from his lawyers, Heppner turned to Claude, Anthropic's generative AI platform. He used it to help outline possible defense strategies.

Heppner gave the AI information he had learned from his lawyers. He asked it to help him prepare reports. The reports laid out defense strategies. They also made arguments about the facts and law he expected the government might charge. He later shared these AI-made documents with his legal team. He believed they would be protected under attorney-client privilege and the work product doctrine.

The government had other ideas. The FBI had seized these "AI Documents" when Heppner was arrested. Prosecutors asked the court to rule that no legal privilege protected them. Heppner's legal team fought back. Among other arguments, they said Heppner made the documents in anticipation of legal proceedings and later shared them with counsel.

Judge Rakoff disagreed entirely.

The Court's Reasoning: No Confidentiality, No Protection

Judge Rakoff's February 17, 2026 opinion reached two important conclusions. It is a single federal district court ruling, so it does not bind other courts. But other judges may find its reasoning persuasive. First, he ruled that attorney-client privilege didn't apply. Claude is not a lawyer, and Heppner's talks with it lacked confidentiality. Second, the work product doctrine gave no protection. Heppner acted without attorney direction.

The Confidentiality Problem

The confidentiality issue proved fatal to Heppner's privilege claims. Judge Rakoff looked at Claude's privacy policy. Users must accept it to use the platform. The policy clearly reserves Anthropic's right to disclose user data to "third parties," including "governmental regulatory authorities." It also lets Anthropic collect what users type and how Claude responds, for training purposes.

The court reasoned that this policy language clearly put users on notice. It told them Anthropic "may" disclose data to third parties. Judge Rakoff also followed another recent decision from the same district. He noted that AI users "do not have substantial privacy interests" in certain conversations. These are conversations they choose to share with an AI platform that keeps those conversations as part of its normal business.

The court compared Heppner's AI documents with private notes a client might write to share with an attorney. It found them different. The key difference: "Heppner first shared the equivalent of his notes with a third-party, Claude."

The Work Product Doctrine Fails Too

The work product doctrine protects materials that parties or their representatives prepare in anticipation of litigation. But this protection usually requires an attorney to direct the work or take part in it. Heppner made his AI documents fully on his own. Counsel never suggested it. So the doctrine offered no refuge.

Judge Rakoff stressed that the documents didn't reflect counsel's strategy at the time they were made. Instead, they showed Heppner's own efforts to analyze his legal situation with help from AI.

Implications for Arizona Legal Practice

Heppner was decided by a federal district court in New York. It is not binding in Arizona. But its reasoning is a useful warning for Arizona attorneys and clients. The decision raises several important points for Arizona attorneys and their clients to consider.

Privacy Policy Pitfalls

Most AI platforms today have similar data-sharing terms in their terms of service. Arizona lawyers who use AI tools for case strategy, document review, or client messages should read these policies closely. Some platforms keep user data or reserve the right to share it with third parties. Those platforms may put at risk a lawyer's duty of client confidentiality under the Arizona Rules of Professional Conduct.

The Supervision Requirement

The Heppner decision suggests that attorney direction and supervision might preserve some protections for AI-assisted work product. Arizona attorneys thinking about adding AI should set clear rules. Those rules should make sure a lawyer oversees any AI use that involves client matters or litigation strategy.

This approach fits existing Arizona State Bar guidance. That guidance encourages lawyers to understand and supervise the technology tools they use in their practice. The key line falls between two situations. In one, clients use AI on their own. In the other, attorneys direct AI use as part of legal representation.

Practical Applications

Injury lawyers in Arizona often handle cases with complex medical records, accident reconstruction, and insurance messages. AI tools now help more and more with document analysis, medical record summaries, and legal research. The Heppner decision stresses why attorneys need to stay in control of these processes. That control helps preserve work product protections.

Criminal defense attorneys, too, must take special care with client messages that involve AI platforms. Arizona has broad discovery rules in criminal cases. Under those rules, prosecutors could get to see AI-made defense materials if they lack proper privilege protection.

The Future of AI in Legal Practice

The Heppner decision doesn't ban AI use in legal practice. Instead, it sets guardrails for keeping traditional privilege protections. Smart lawyers will adapt. They will use AI tools under proper attorney supervision. They will also avoid platforms that put client confidentiality at risk.

Arizona's legal community should expect more guidance as courts face more AI-related privilege questions. The Arizona Supreme Court's Ethics Advisory Committee may issue formal opinions. Those opinions could clarify ethical duties when lawyers use AI tools to represent clients. The State Bar of Arizona also provides practical guidance and best practices documents on this topic.

Meanwhile, AI developers are likely to build more platforms made just for legal use, with a stronger focus on privacy. These may include features that better protect attorney-client privilege and work product.

Frequently Asked Questions

Does the Heppner decision ban lawyers from using AI tools?

No, the Heppner decision doesn't ban AI use in legal practice. The court held that privilege did not cover Heppner's AI chats because Claude is not a lawyer and the chats were not confidential, and work product protection failed because he acted without attorney direction. But attorneys can still use AI tools under proper supervision and with fitting privacy safeguards. The key is for the attorney to stay in control of the AI interaction. The attorney must also make sure the platform's privacy policy doesn't put client confidentiality at risk.

How does this ruling affect Arizona attorneys specifically?

Arizona attorneys must now look closely at how AI platforms handle privacy and store data. They must do this before using them for client matters. The decision reinforces existing ethical duties under the Arizona Rules of Professional Conduct. Those duties cover client confidentiality and competent representation. Arizona lawyers should put clear rules in place for supervising AI. They should also avoid platforms that share data with third parties.

What should clients know about using AI for legal advice?

Using AI platforms alone, without a lawyer's direction, to analyze a legal situation or plan a defense carries real risk. Under the reasoning in Heppner, those conversations may not be protected. Opposing parties or prosecutors could obtain them in discovery. Instead, clients should consult qualified attorneys. Those attorneys can properly supervise any AI-assisted legal work while keeping proper confidentiality protections in place.