Earlier this year, the Honorable Judge Rakoff, U.S.D.J., sitting in the United States Southern District of New York, issued a ruling in a criminal case (United States of America v. Heppner, U.S. D. Ct. Southern District of N.Y, 25 Cr. 503) that should cause litigants, or potential litigants, to think twice before using an AI tool to strategize.

CASE SUMMARY

In the Heppner case, Mr. Heppner was indicted by a grand jury on securities fraud, wire fraud…charges. Agents from the F.B.I. executed a search warrant and seized documents and electronic devices, including approximately 31 documents reflecting communications that Mr. Heppner allegedly had with Anthropic’s “Claude” AI platform. The documents included defense strategies. Mr. Heppner’s counsel claimed the Claude queries included information Mr. Heppner learned from his attorneys, the AI documents were generated for the purpose of speaking with counsel to obtain legal advice, and the contents of the AI-generated documents were shared with counsel.

The government filed a motion seeking an order that the AI documents are neither protected by the attorney-client privilege nor are they covered by the work product doctrine. The government prevailed.

The Court determined:

I. The attorney-client privilege does not apply because:

  • The communications with AI were not between Mr. Heppner and his attorney.
  • Also, the communications with AI were not confidential. The Court noted that Claude’s use of terms reflects that input and output are used to train the AI model, and it reserves the right to disclose the information to third parties.
  • Finally, Mr. Heppner did not communicate with Claude for the purpose of obtaining legal advice, nor did he conduct the AI research at the direction of counsel. Plus, Claude’s output disclaims that it is legal advice.

II. The work product doctrine does not apply since the defendant’s AI queries were not materials prepared by or at the request of counsel in anticipation of litigation.

Significance

While the Heppner case was decided in the context of a criminal proceeding, the ruling’s implications are likely to extend to civil matters. For example, a party that conducts AI-assisted analysis, evaluation… without the advice and at the direction of counsel runs the risk that the results may not be covered by the privileges discussed above and may end up being used against the party.

In addition, uploading sensitive or confidential information into an AI tool may result in the information being unwittingly added to the tool’s model, not only jeopardizing your position in a negotiating or litigation context but also resulting in additional claims. For example, if the information submitted to the AI tool was provided under a non-disclosure agreement (“NDA”), by uploading the information to the tool, you could be violating the NDA.

Also, check the use terms for the AI tools that you use, to determine what they do with your inputs and outputs, including whether their models train on this information, and whether they keep it confidential…

The Heppner case reinforces the need for businesses to have in place an AI use policy that includes provisions that address AI use for legal matters, restricts uploading confidential information, and provides employee training on the use of AI.

If you have questions about AI and its potential impact on your business, please contact Angelo Bolcato, Esq. at Laddey Clark & Ryan, LLP at [email protected], or (973) 729-1880.

The foregoing is not intended, and should not be construed as, legal advice or guidance or an offer to provide legal services by Laddey, Clark & Ryan, LLP. The contents of this communication are for informational purposes only and should not be relied on, or considered, in making any decisions or taking any actions. If you wish to inquire about legal services, please contact Laddey, Clark & Ryan, LLP, at [email protected].