SUMMARY
If you have been treating your AI tool as a legal advisor, you are courting trouble beyond the well-documented risks of hallucinations and invented citations.
A federal court in the Southern District of New York has now ruled that neither the information you share with an AI platform nor the output it generates will be treated as confidential under the doctrine of attorney-client privilege, which traditionally shields all communications between lawyers and their clients.
In the first federal ruling to directly address privilege claims over AI-generated materials, Judge Jed S. Rakoff held that 31 documents created by defendant Bradley Heppner using Anthropic’s Claude were neither protected by attorney-client privilege nor the work product doctrine.
Key points:
- Consumer AI tools are not attorneys and owe no duty of confidentiality; documents created through them do not attract attorney-client privilege
- Independent, client-driven AI use without attorney direction does not qualify for work product protection
- Transmitting AI-generated documents to counsel after the fact does not retroactively confer privilege
- Inputting attorney-sourced information into a consumer AI platform constitutes disclosure to a third party and waives any existing privilege
- Courts will apply traditional privilege rules to AI-generated materials; no new protections will be created for technology
INTRODUCTION
On February 10, 2026, Judge Jed S. Rakoff ruled in United States of America v. Bradley Heppner that approximately 31 documents generated by Heppner using Anthropic’s Claude were neither protected by attorney-client privilege nor the work product doctrine, making it the first federal ruling to directly address privilege claims over consumer AI-generated materials.
The central question was whether documents independently created by a defendant using a commercial AI tool, after receiving information from his attorneys and intending to discuss the output with counsel, could qualify for legal protection.
The court answered in the negative, drawing a clear line between unstructured, client-driven AI use and scenarios where privilege may be preserved, such as AI use directed by counsel or conducted through enterprise platforms with enforceable confidentiality guarantees.
BACKGROUND
The Charged Conduct
Heppner was indicted on October 28, 2025, and arrested on November 4, 2025, on charges of securities fraud, wire fraud, conspiracy to commit securities and wire fraud, making false statements to auditors, and falsification of records.
The charges arose from an alleged scheme to defraud investors in Beneficient, a financial services company Heppner founded and led as CEO.
The indictment alleged that he misrepresented a debt owed by Beneficient to Highland Consolidated Limited Partnership (HCLP), portraying it as an independent lender associated with a wealthy family, while in reality secretly controlling HCLP and personally benefiting from funds transferred to it.
As Chairman of GWG Holdings, he directed payments to Beneficient ostensibly to repay that debt, allegedly diverting over $150 million for personal use, including renovations to his Dallas mansion and private travel expenses.
The AI-Generated Documents
Following receipt of a grand jury subpoena but before his arrest, Heppner used Claude to create documents comprising his prompts and AI-generated responses, incorporating factual and legal information obtained from defense counsel.
He did so on his own initiative and without attorney direction, intending to use the output as a basis for discussions with counsel, and subsequently transmitted the documents to his lawyers.
Federal agents seized the documents during a search of his residence, after which defense counsel asserted privilege and the government moved for a ruling that they were not protected.
THE APPLICABLE LEGAL FRAMEWORK
Attorney-Client Privilege
Attorney-client privilege protects confidential communications between a client and their attorney made for the purpose of obtaining or providing legal advice. As established in United States v. Mejia, it applies only where the predominant purpose is to render or solicit legal advice and is narrowly construed as it impedes the search for truth.
The communication must involve a licensed attorney, be made in confidence, and not be disclosed to third parties. Voluntary disclosure to a party owing no duty of confidentiality constitutes a waiver.
In evolving legal environments, businesses often consult with artificial Intelligence lawyers to understand how emerging technologies intersect with traditional privilege doctrines.
The Work Product Doctrine
The work product doctrine provides qualified protection for materials prepared by or at the direction of counsel in anticipation of litigation.
As articulated in In re Grand Jury Subpoenas Dated Mar. 19, 2002 and Aug. 2, 2002, it does not extend to documents created independently by a layperson without attorney direction, and requires that materials reflect an attorney’s thought processes or have been compiled under attorney supervision.
Complex regulatory and digital frameworks are frequently navigated with the support of a technology law firm experienced in data governance and AI compliance.
THE COURT’S RULING
The court held that the AI-generated documents failed every element of both tests, resting its decision on three independent grounds, each individually sufficient to defeat the privilege claim.
First, Claude is not a licensed lawyer, holds no bar admission, and owes no duties of loyalty or confidentiality to users.
Second, Heppner could not credibly claim to have sought legal advice from the platform, a conclusion reinforced by Claude’s own terms of service, which disclaim any attorney-client relationship and direct users to consult qualified attorneys.
Third, there was no confidentiality: Claude is a publicly accessible consumer platform whose privacy policy permits data use for training and disclosure to authorities, leaving no reasonable expectation of privacy over information entered by users.
The work product doctrine fared no better. Heppner created the documents on his own initiative without attorney direction, and the doctrine does not extend to a client’s independent AI research. The court rejected the argument that transmitting the documents to counsel after the fact brought them within its protection.
On waiver, relying on United States v. Ackert and In re von Bulow, the court held that by inputting attorney-sourced information into a third-party platform, Heppner effectively disclosed it to a third party, waiving any privilege that might otherwise have attached.
The ruling distinguished this from scenarios where protection may survive, such as AI use directed by counsel or conducted through enterprise platforms with contractual confidentiality guarantees. Businesses addressing such risks often rely on experienced corporate lawyers to structure compliant legal and technology strategies.
Authors: Shantanu Mukherjee, Alan Baiju
