Client’s Use of AI Not Protected by Attorney Client Privilege

Published
Score
3

Why it matters

In United States v. Heppner, a federal judge in the Southern District of New York (SDNY) ruled that 31 documents generated by defendant Michael Heppner using Anthropic's consumer Claude AI—containing details from his counsel and defense strategy outlines related to fraud charges—were not protected by attorney-client privilege or the work-product doctrine.[1][2][3][4][7] The government seized these documents from Heppner's devices, prompting a dispute; Judge Jed S. Rakoff ruled from the bench on February 10, 2026, with a written opinion issued on February 17, 2026, marking the first federal decision on privilege for consumer AI outputs.[1][3][4][6][7]

Heppner, facing criminal fraud charges, independently inputted privileged information from his lawyers into Claude to create documents for legal advice, then shared them with counsel; the government argued against privilege, citing: (1) no attorney-client relationship, as Claude is not a lawyer; (2) no reasonable expectation of confidentiality due to Anthropic's privacy policy allowing data use for training and third-party disclosure; and (3) Heppner acted on his own volition, not at counsel's direction.[1][2][3][4][5][6] Involved parties include Heppner, his unnamed attorney, the U.S. government (prosecutors), Anthropic (Claude provider), and Judge Rakoff; no specific companies beyond Anthropic are named as defendants.[2][4][7]

The case arose amid rising AI use in legal contexts, highlighting risks when clients input privileged data into public AI tools without safeguards; courts emphasized that privilege requires human attorney relationships, confidentiality, and intent for legal advice—none met here.[2][5][6] Timeline: AI use predates seizure; motion argued pre-February 10 ruling; decision publicized February 20, 2026.[3][4][7]

Newsworthy now as the first-of-its-kind federal precedent amid explosive generative AI adoption, warning HR, legal teams, and companies against consumer AI in sensitive matters like investigations—potentially waiving privilege via third-party disclosures—urging shifts to secure, counsel-directed enterprise tools.[1][2][5][6][9]

Sources

mail

Get notified about new Privacy developments

Primary sources. No fluff. Straight to your inbox.

See more entries tagged Privacy.