CLIENT ALERT: Southern District of New York Judge Rakoff Rules that Defendant Accused of Fraud Cannot Assert Privilege Over AI-Generated Documents

KEY TAKEAWAYS

• On February 10, 2026, Judge Jed. S. Rakoff of the Southern District of New York found that documents a defendant created using an AI tool and subsequently sent to his attorneys were not protected by the attorney-client privilege or attorney work-product privilege.

 

• Judge Rakoff noted that while the relevant materials were not privileged, the prosecution’s use of them could create a “witness-advocate conflict” at trial and potentially lead to a mistrial.

 

• When a client creates a document using an AI tool and subsequently shares it with counsel, it is not automatically privileged, particularly when the document does not reflect legal advice from the client’s counsel or the AI platform the client used did not have confidentiality protections.

 

Factual BACKGROUND

In November 2025, Bradley Heppner, founder of Beneficient, an alternative finance firm, was indicted in the Southern District of New York on charges of securities fraud, wire fraud, conspiracy to commit securities and wire fraud, making false statements to auditors, and falsification of records. The indictment accused Heppner of funneling hundreds of millions of dollars in investments from GWG Holdings, Inc., a financial services company whose board he chaired, into Beneficent to pay off debts it owed to an entity that turned out to be a shell company secretly controlled by Heppner.

Defense counsel informed the Government that, prior to his arrest, Heppner had used Anthropic’s AI “assistant” tool, known as “Claude,” to conduct queries related to the Government’s investigation, creating approximately 31 documents reflecting those queries and Claude’s responses (the “AI Documents”). Defense counsel argued that the AI Documents were privileged because Heppner had incorporated information he had learned from his attorneys into his AI queries, created the AI Documents to obtain legal advice, and subsequently shared them with defense counsel by email. However, counsel acknowledged that they did not direct the defendant to use the AI tool and were not otherwise involved in creating the AI Documents.

The Government argued that the AI Documents were not protected by either the attorney-client privilege or the work-product doctrine.

The Court’s Holding

In an oral ruling on February 10, 2026, the Court determined that there was no basis to treat the AI Documents as privileged because they were prepared by the defendant himself in collaboration with the AI platform and could not be considered communications seeking legal advice from counsel.

First, the Court found no basis for attorney-client privilege despite the defendant sharing the AI documents via email with defense counsel. This is consistent with well-settled precedent in the Southern District that non-privileged or preexisting documents do not become privileged solely through transmission to counsel.

Second, the Court agreed with the Government’s argument that the documents were not protected by the work-product doctrine because the AI Documents did not reflect defense counsel’s legal strategy and were instead prepared by an AI platform. The Court also noted that Claude’s terms indicated that information entered into the platform was not confidential, thereby undermining any claim of privilege.

Although the Court declined to extend privilege to the AI Documents, when Heppner’s attorney flagged that their admission at trial could create a witness-advocate conflict, whereby defense counsel would be called as a witness regarding the documents, the Court acknowledged the possibility.

Practical Implications

Even if represented by counsel, clients should not assume that materials created using AI tools or platforms will be protected by attorney-client privilege or work-product doctrine, even if they later share those materials with counsel to aid in their representation. In practice, this means that drafting documents using, or asking for legal advice from, third-party AI platforms may create discoverable evidence that can be used in litigation. Individuals and entities should therefore consult counsel before using AI tools or platforms for sensitive matters and review the confidentiality terms of the platforms if they decide to use them. Corporate entities should adopt internal policies to limit the entry of legal or investigative information into non-secure systems. Where the relevant matter is already before a court, counsel should also review the applicable local and individual court rules, given many courts have begun adopting new rules regarding litigants’ use of AI.

In the event that a client does create discoverable material using AI, counsel should consider whether use of the material would implicate a witness-advocate conflict potentially rendering the material unusable at trial.

CONCLUSION

The Court’s ruling underscores that materials a client independently generates using AI platforms or tools may fall outside of attorney-client privilege and work-product doctrine. The Court’s ruling serves as a caution to clients and organizations to treat AI platform input as potentially discoverable and to communicate with counsel regarding the use of AI during any government investigation or litigation. Sher Tremonte LLP will continue to monitor developments. For questions, please contact us.

Written by Brian Kidd, Taylor Fontan, and Rebecca Prager.