Key Takeaways
As use of artificial intelligence (“AI”) grows, it is increasingly important to understand the pros and cons of its use.
As made apparent by a recent decision of the United States District Court for the Southern District of New York (“SDNY”), use of publicly available AI platforms may result in disclosure of confidential information and information believed to be privileged.
To manage this risk, it is essential to carefully review the terms of use and service and privacy policies issued by AI systems you employ, to educate your employees on and monitor use of AI systems as appropriate, and to consider restrictions on AI use as necessary and appropriate to preserve privilege and confidentiality.
Introduction
As use of AI becomes more frequent and ever present in how businesses operate, organize information, communicate, and boost efficiencies, it is vital to understand the risks involved—particularly in the context of legal representation.
A recent decision by SDNY highlights the danger that the use of publicly available AI poses to attorney-client privilege and confidentiality, and showcases courts’ willingness to scrutinize clients’ use of AI in the course of obtaining legal advice.
A brief overview of the attorney-client privilege, SDNY’s decision in United States v. Heppner, No. 25-CR-000503 (S.D.N.Y. Feb. 17, 2026) (“Heppner”), and a breakdown of the “dos” and “don’ts” associated with AI use in the wake of Heppner, is discussed below.
Attorney-Client Privilege & Confidentiality
Attorney-client privilege shields from disclosure confidential communications between a client and the client’s attorney in connection with obtaining or rendering legal advice. It is designed to promote the disclosure of full, honest, and complete information by a client to their attorney in exchange for competent advice.
The privilege, however, is not absolute and may be waived through disclosure of a privileged communication to a third party other than the client, their attorney, or their attorney’s agent (in select instances).
As further discussed below, SDNY’s decision in Heppner clarifies that publicly available AI platforms constitute “third parties” for purposes of destroying confidentiality and waiving existing privilege.
Overview of SDNY’s Heppner Decision
Although Heppner was a criminal case, its principles apply to the civil context with equal force. In Heppner, the client disclosed information he learned from his attorney to Claude, the AI chatbot built by Anthropic. In concluding that such disclosure was not protected by the attorney-client privilege, SDNY reasoned that (1) Claude was a third party, disclosure to which would not have been confidential and/or would have waived privilege; (2) Claude was not a human capable of providing legal advice and, indeed, disclaimed its ability to do so in its terms; and (3) Claude’s privacy terms warned that its information inputs and outputs may not only be used to train Claude but also may be disclosed to third parties, such as governmental regulatory authorities.
SDNY also found that documents created by Claude in connection with the client seeking legal advice from his attorney were neither privileged—despite later disclosure to the attorney—nor protected under the work product doctrine, which, at a high level, protects documents created by or at the behest of an attorney in anticipation of litigation.
Takeaways from the Heppner Decision
In the wake of Heppner, businesses should be particularly vigilant, cautious, and restrictive in their use of publicly available AI systems that use information inputs and outputs for machine learning or training and that retain the right to disclose such information to third parties—i.e., systems that afford users no expectation of privacy. Heppner makes clear that such systems’ inputs and outputs related to obtaining legal advice from an attorney or that disclose attorney-client communications are non-confidential and non-privileged.
It is not only disclosures to AI made prior to or following consultations with an attorney that pose risk here, however. Of arguably even more danger are AI systems employed during meetings or calls with counsel, such as add-ins to Zoom or Microsoft Teams meetings to record conversations for transcription, note-taking, and summarizing purposes. Applying Heppner, the presence of a public AI system during such consultations would qualify as a “third party,” breaching the protection that confidentiality and privilege affords.
The “Dos” and “Don’ts” of AI Use
-
Do:
- Carefully study the terms and privacy policies of AI systems to understand whether the information put into and generated by the system may be used for machine learning, shared with third parties, or otherwise publicly available, before using them;
- Understand the AI systems’ capabilities and limitations; note-taking or summarizing systems, for example, not only pose risk to confidentiality and privilege, but also may fail to accurately and completely capture the conversation’s content and tone;
- Educate employees on the risks AI use poses to confidentiality and attorney-client privilege;
- Consider restrictions, as appropriate, on AI use to protect and preserve attorney-client confidentiality and privilege; and
- Understand how to disable or block AI systems’ access to confidential and attorney-client communications, as appropriate.
-
Don’t:
- Input otherwise confidential and privileged communications into publicly available AI systems, or allow access to the same by AI systems, as this likely waives any existing privilege or protection;
- Use AI to generate summaries, analyses, documents, or correspondence in connection with seeking legal advice. Later disclosing the information input into or generated by AI to your attorney does not cloak such information in confidentiality or privilege; and
- Invite note-taking or other AI systems to join confidential consultations with your attorney, without disclosing the tool and discussing its risks with your attorney first.
Conclusion
As SDNY recognized in Heppner, “[g]enerative [AI] presents a new frontier in the ongoing dialogue between technology and the law.” The attorneys at Smith Currie Oles strive to keep you informed on the legal landscape of this new frontier. If you have questions regarding the use of AI and any implications for you or your business, please contact one of our experienced attorneys.
Smith Currie Oles LLP provides comprehensive legal services to all parts of the construction industry across the nation. Smith Currie lawyers have decades of demonstrated success representing construction and federal government contracting clients “From the Ground Up,” including procurement matters, contract formation and negotiation, project administration, claims prosecution and, when necessary, in litigation and other forms of dispute resolution.
The views expressed in this article are not necessarily those of ConsensusDocs. Readers should not take or refrain from taking any action based on any information without first seeking legal advice.