Stephanie Messas, Esq., Alastair Mecke, Esq., and Noémie Tabiou
The increasing integration of artificial intelligence (AI) into legal practice offers undeniable efficiency gains. However, it also raises significant legal concerns, particularly with respect to attorney-client privilege and the protection of confidential information. These risks are not merely theoretical. Recent judicial developments suggest that improper use of AI tools may result in the inadvertent waiver of privilege.
A Brief Overview of Attorney-Client Privilege
Attorney-client privilege is a legal protection that keeps communications between a client and their lawyer confidential, provided those communications were made for the purpose of obtaining legal advice.
This protection prevents disclosure to the government or opposing parties. However, the privilege only works if the communication was kept confidential in the first place. Once confidentiality is compromised, the privilege may be lost.
The United States v. Hepner Decision
This risk was highlighted in United States v. Hepner (S.D.N.Y., Feb. 17, 2026). In that case, the defendant, facing federal criminal charges, used a publicly available generative AI platform on his own initiative to analyze his legal exposure and possible defense strategies. He later shared the AI‑generated materials with his attorneys.
Federal agents later seized his electronic devices and found not just the AI-generated documents, but also the full AI prompt history. His defense team claimed these materials were protected by attorney-client privilege, describing them in a legal filing as “artificial intelligence-generated analysis conveying facts to counsel for the purpose of obtaining legal advice.” The court disagreed.
Judge Jed S. Rakoff held that the defendant’s use of a consumer-grade AI platform destroyed confidentiality and therefore prevented the application of attorney-client privilege or work product protection. The court identified that Hepner’s use of the AI platform broke confidentiality in three distinct ways:
- Use of a Public AI Platform: The AI provider’s terms allowed data collection and use. By inputting sensitive legal information into the platform, the defendant disclosed it to a third party, defeating confidentiality.
- AI Is Not a Privileged “Agent”: The court rejected the argument that the AI tool qualified as a protected third-party assistant under the Kovel doctrine (which allows privilege to extend to certain non-attorney professionals, like translators or accountants, who assist attorneys in providing legal advice). Unlike an accountant or translator acting at an attorney’s direction, the AI was used independently and without attorney supervision.
- No Retroactive Privilege: Privilege must exist at the time of communication. Sharing information with a third party and later providing it to counsel does not restore lost confidentiality.
The court also declined to apply the work product doctrine (a separate protection that shields materials prepared by or for an attorney in anticipation of litigation), noting that the documents reflected the defendant’s own thinking and that any protection was waived once the information was shared with the AI platform.
What This Decision Does—and Does Not—Mean
This decision does not prohibit the use of AI in legal practice. The ruling was highly fact-specific and issued by a federal district court, meaning it is not binding on other courts. Judge Rakoff himself noted that different facts, such as attorney-directed use of secure AI tools, could potentially lead to a different result.
That said, the case serves as a clear warning about the risks of using unsecured or consumer‑grade AI tools for legal analysis.
An Important Warning for Clients
Clients should exercise extreme caution when using AI tools in connection with legal issues. If you wish to keep information confidential, you should not input it into publicly available or consumer AI platforms. This includes facts about potential disputes, internal investigations, business strategies, or communications related to ongoing or anticipated litigation.
Using AI to analyze legal problems before speaking with counsel—or without your attorney’s knowledge—can unintentionally waive important legal protections. Once confidentiality is lost, it often cannot be restored.
If you are considering using AI in connection with a legal matter, we strongly recommend discussing it with your attorney first. We can help assess whether a particular tool or approach is appropriate and how to protect privileged information.
Conclusion
AI can be a valuable tool, but it must be used carefully and strategically. The Hepner decision underscores a simple but critical principle: confidentiality is the foundation of attorney‑client privilege. Clients and practitioners alike should remain vigilant to ensure that new technologies do not undermine long‑standing legal protections.