In a groundbreaking February 2026 ruling, Judge Jed Rakoff of the Southern District of New York held that communications with consumer AI tools—such as the publicly available version of Anthropic’s Claude—are not protected by attorney‑client privilege or the work product doctrine.
That means your opponent in a civil lawsuit over a contract breach, or the government in a criminal case charging you with a crime, can request that you provide all communications you had with publicly available Artificial Intelligence Large Language Models (LLMs), including legal strategies, defenses and the like. And the court may well order that you produce those communications.
This decision in United States v. Benjamin Heppner, 25 Cr. 503 in New York is the first federal opinion to directly address whether using generative AI in connection with legal matters can create privileged communications. The court’s answer: no.
Such a ruling could have major implications for law firms, in‑house counsel, compliance teams, and any organization using AI tools for legal research or litigation strategy.
Why This Case Matters for AI, Privilege, and Legal Risk
Defendant Bradley Heppner used the consumer version of Claude to:
-analyze factual and legal issues related to a potential indictment,
-incorporate information learned from his attorneys, and
-generate strategy documents he later shared with counsel.
Heppner argued these “AI Documents” were privileged because they were created to obtain legal advice.
Judge Rakoff disagreed.
The Court reasoned that the client’s communications with Claude AI were able to be obtained by the client’s opponent, because:
1) AI Is Not an Attorney — and Cannot Create Privileged Communications
Privilege requires a communication between a client and a lawyer.
2. Users Have No Reasonable Expectation of Confidentiality With Consumer AI Tools.
That is because the consumer version of Claude trains on user inputs, Anthropic’s privacy policy allows disclosure to “third parties,” including government authorities, and the policy warns that data may be shared “in connection with claims, disputes, or litigation.”
3. The client used AI on his own, which does not fall under an attorney’s work product doctrine. That doctrine protects materials prepared by or at the direction of counsel that reflect attorney strategy or the attorney’s mental impressions.
The maxim holds true: If the service is free, then you are the product. Google reads your Gmail emails to teach its AI LLM. Claude learns from user inputs to increase the effectiveness of its responses. CoPilot compiles and relies on user inputs to get better, faster, and smarter. Facebook and Meta use your photos and private DMs (direct messages) to improve its AI. Even if they were private at one time, they are no longer. Company policies change. Have you read the latest user agreement for these free services you use everyday?
Most privilege waivers happen unintentionally.
Conclusion: A Landmark Decision With Potential National Implications
Judge Rakoff’s ruling is the first federal decision to directly answer whether communications with a publicly available AI platform can be privileged. His answer is clear:
Communications with consumer AI tools are not privileged and not protected work product. Organizations should immediately review their AI usage policies, especially in litigation, investigations, and sensitive internal matters.
Call Savarese & Associates today for a consultation. 228-901-0000
