Please ensure Javascript is enabled for purposes of website accessibility

Client Alert

In a decision with significant implications for any non-lawyer who uses artificial intelligence tools to research or analyze legal matters, Judge Rakoff of the United States District Court for the Southern District of New York, in United States of America v. Heppner, 25-cr-00503-JSR, ruled on February 10, 2026 that documents generated through a public AI platform were not protected by the attorney-client privilege or the work product doctrine.1 Specifically, the Court granted the Government's motion to access documents that defendant Bradley Heppner created using the AI tool Claude before his arrest on federal fraud charges.

Background

Heppner is charged with securities fraud, wire fraud, conspiracy, making false statements to auditors, and falsification of records arising from an alleged scheme to defraud investors in connection with a financial services company he founded and controlled as CEO.2

During the execution of a search warrant at Heppner’s residence, federal agents seized electronic devices containing documents that Heppner had generated by inputting queries into Claude relating to the Government’s investigation (the “AI Documents”).3 Heppner’s counsel informed the government that Heppner created the AI Documents before his arrest and later shared them with his attorneys, and that some of the information Heppner input into Claude he had learned from his counsel.4 Heppner asserted that the AI Documents were privileged and requested their segregation from the Government attorneys, a request to which the Government agreed pending a privilege determination by the Court.5

The Government’s Motion

The Government filed a “Motion for a Ruling that Documents the Defendant Generated Through an Artificial Intelligence Tool Are Not Privileged” in which it argued that the AI Documents were not protected by the
attorney-client privilege and not shielded by the work product doctrine.6

The attorney-client privilege “protects communications (1) between a client and his or her attorney (2) that are intended to be, and in fact were, kept confidential (3) for the purpose of obtaining or providing legal advice.”7 The Government made a number of arguments as to why the attorney-client privilege did not apply to the AI Documents:

  • The AI Documents were not communications between Heppner and his counsel. Rather, they were made between Heppner and a public AI tool that is not an attorney. The Government analogized Heppner’s use of Claude to him asking friends for input on legal matters, an act that does not create privileged communications.8
  • The AI Documents were not created for the purpose of obtaining legal advice from counsel. Specifically, Claude’s “Constitution,” terms of service, and other public materials expressly disclaim Claude’s ability to give legal advice and instead suggest that the user consult with a qualified lawyer.9
  • The AI Documents were not confidential. Rather, Heppner voluntarily shared his prompts with a third-party commercial AI platform that is publicly accessible. Claude’s Privacy Policy explicitly advises users that it collects data on prompts and outputs, uses this data to train its AI tool, and may disclose this data to governmental regulatory authorities and third parties.10
  • Heppner’s subsequent transmission of the AI Documents to his attorneys did not retroactively cloak them with privilege. Well-settled law holds that sending non-privileged documents to counsel does not make them privileged.11 It is equally well-established that while a communication to one’s attorney may be privileged, the underlying information or facts in that communication are not.12

The Government also argued that the AI Documents were not shielded by the work-product privilege, which “provides qualified protection for materials prepared by or at the behest of counsel in anticipation of litigation or for trial.”13 Heppner admittedly created the AI Documents on his own initiative, not at counsel's direction, and thus should not be able to later claim that he prepared the documents at the behest of counsel in anticipation of trial, as required to assert the work product privilege.

The Court’s Ruling

The Court orally granted the Government's motion from the bench on February 10, 2026, holding that Heppner failed to meet his burden of establishing both attorney-client privilege and the work product privilege over the AI Documents.14

Takeaways and Implications

While this ruling is based on fundamental and well-established rules regarding attorney-client privilege and the work product doctrine, and thus is not necessarily surprising, it nevertheless has potentially far-reaching consequences for non-lawyer individuals and organizations who use public AI tools to research legal matters:

  • Communications with public AI tools may not be privileged. Communications with public AI tools may not satisfy the requirements of the attorney-client privilege because AI tools are not attorneys, do not provide legal advice, and inputs to them are not confidential. Further, later sending the AI-generated results to a lawyer will not retroactively cloak them with privilege.
  • Public AI tool privacy policies are important. Courts may examine whether AI tool privacy policies permit disclosure of user data to third parties and governmental authorities when evaluating confidentiality claims so such policies should be carefully reviewed prior to use of the AI tool.
  • Sharing privileged information with a public AI tool might waive privilege. Taking the ruling a step further, it is reasonable to also conclude that sharing confidential attorney-client communications with a public AI tool might waive any privilege that could otherwise attach to those communications.

Clients and counsel should carefully consider these implications before using public AI tools to analyze confidential legal matters. Further, organizations may want to review their policies and procedures regarding the use of public AI tools in connection with legal matters, particularly in the context of ongoing or anticipated litigation.


  1. Minute Entry, USA v. Heppner (S.D.N.Y. Feb. 10, 2026).
  2. Redacted Indictment, Heppner (S.D.N.Y. Oct. 28, 2025), ECF #3.
  3. Motion for a Ruling that Documents the Defendant Generated Through an Artificial Intelligence Tool Are Not Privileged at 3, Heppner (S.D.N.Y. Feb. 6, 2026), ECF #22.
  4. Id. at 3-4.
  5. Id. at 4.
  6. Id.
  7. United States v. Mejia, 655 F.3d 126, 132 (2d Cir. 2011).
  8. Motion at 7, Heppner (S.D.N.Y. Feb. 6, 2026).
  9. Id. at 8.
  10. Id. at 11.
  11. United States v. Correia, 468 F.Supp.3d 618, 622 (S.D.N.Y. 2020).
  12. Upjohn Co. v. United States, 449 U.S. 383, 395 (1981).
  13. In re Grand Jury Subpoenas Dated Mar. 19, 2002 and Aug. 2, 2002, 318 F.3d 379, 383 (2d Cir. 2003).
  14. Minute Entry, Heppner (S.D.N.Y. Feb. 10, 2026).

Related People

We have always been focused on finance.

  • 1913
    TS Chapman partners with Henry Cutler to form Chapman and Cutler
  • 1st
    Chapman's first client in 1913 is still a client of the firm today
  • 22
    Diverse financial practices serving regional, national, and global clients
  • 6
    Offices across the country and in key US financial centers

Chapman and Cutler LLP Cookie Preference Center

Your Privacy

When you visit our website, we use cookies on your browser to collect information. The information collected might relate to you, your preferences, or your device, and is mostly used to make the site work as you expect it to and to provide a more personalized web experience. For more information about how we use Cookies, please see our Privacy Policy.

Strictly Necessary Cookies

Always Active

Strictly Necessary cookies enable core functionality such as security, network management, and accessibility. These cookies may only be disabled by changing your browser settings, but this may affect how the website functions. Some functions of the site require remembering user choices, for example your cookie preference, or keyword search highlighting. When using a contact form or event registration, a cookie might be used to monitor the state of your submission across pages. These do not store any personal information.

Marketing Cookies

Marketing cookies help us improve our website by collecting and reporting information on its usage. We access and process information from these cookies at an aggregate level.

Powered by Firmseek