Guide

Are your AI chats privileged? What the courts said in 2026

By Nick Steiner, founder of Confidelio · Updated October 21, 2026

Short answer: Usually not. In February 2026 a federal judge in New York ruled that a defendant's conversations with a consumer AI were not protected by attorney-client privilege, because the AI is not a lawyer and its provider's privacy policy allowed disclosure. A separate court protected a self-represented person's AI work as work product. No state has passed a law making AI chats privileged.

What is attorney-client privilege?

Attorney-client privilege is the rule that keeps certain communications with your lawyer out of court. Federal courts in New York describe it as protecting communications between a client and an attorney that are intended to be, and in fact are, kept confidential, for the purpose of getting or giving legal advice. All three parts have to be true.

Sharing a privileged communication with an outsider usually waives the privilege. That is why where your words go matters as much as what you say.

Did a court decide whether AI chats are privileged?

Yes. In United States v. Heppner, Judge Jed S. Rakoff of the Southern District of New York ruled from the bench on February 10, 2026, and explained his reasons in a written memorandum filed February 17, 2026. The court called it a question of first impression nationwide, and the answer was no.

The defendant had used a consumer version of Claude, made by Anthropic, after he received a grand jury subpoena and without any suggestion from his lawyers. The FBI later seized the documents he made with it.

Why weren't the chats privileged?

The court found that the conversations failed the test in at least two ways, and possibly all three:

The AI is not a lawyer. Privilege requires a relationship with an attorney, and the court said that alone disposed of the claim.
The chats were not confidential. The provider's privacy policy allowed it to collect what users type, use it, and disclose it to third parties, including in connection with litigation and to government authorities.
The chats were not a request for legal advice from a lawyer. The AI itself had told the user it was not a lawyer and could not give formal legal advice.

The court added that sending the AI's output to a lawyer afterward does not turn it into a privileged communication, and that typing privileged information into the AI waived the privilege over that information.

Were the chats protected as work product?

No. Work product protects materials prepared in anticipation of litigation, but the Heppner court found these documents were not prepared by or at the direction of his lawyers. The court noted that the result might arguably be different if a lawyer had directed the use of the AI, but it did not decide that question.

Has any court protected someone's AI chats?

Yes, in a narrower way. In Warner v. Gilbarco, decided February 10, 2026, a federal court in Michigan refused to make a self-represented plaintiff hand over her ChatGPT materials. The court held they were her work product, and that using an AI tool did not waive that protection because the AI is a tool, not a person, and nothing had been disclosed to her opponent.

Heppner and Warner do not conflict. Heppner was a criminal case about privilege and work product prepared without a lawyer's direction. Warner was a civil case about whether using AI waives work product that a self-represented party prepared for her own case.

Is there a law that makes AI conversations privileged?

Not yet. Arizona HB 2410, introduced in January 2026, would have treated a person's communications with AI as privileged as if they had been made to a human professional. It passed the Arizona House 53 to 4 in February 2026 but had not become law as of October 2026. We are not aware of any state that has enacted such a law.

OpenAI's chief executive, Sam Altman, said in July 2025 that the industry had not yet figured out legal confidentiality for conversations with ChatGPT.

What does this mean for you?

Assume anything you type into a general-purpose AI about a legal problem could be requested in a lawsuit or criminal case.
Do not delete chats once a dispute is likely. Destroying evidence can be punished.
Tell your lawyer if you have used AI about your matter.
If you want AI help with a legal matter, use a tool designed for confidentiality, and read its privacy policy before you type anything.

Common questions

Are ChatGPT conversations protected by attorney-client privilege?

No court has held that conversations with a general-purpose AI are privileged. The leading 2026 decision, United States v. Heppner, held that chats with a consumer AI were not.

Can my AI chats be used against me in court?

They can be requested in discovery or seized under a warrant like other electronic records. Whether they are admissible depends on the case.

Does sending my AI chat to my lawyer make it privileged?

No. The Heppner court said material that is not privileged does not become privileged by being shared with counsel later.

Legal research that stays yours

Confidelio is a legal research AI designed to keep your work under attorney-client privilege. Your cases are encrypted with a key only your password unlocks, the AI keeps nothing, and nothing you write trains a model.

See how Confidelio works How we protect you

Sources

Related guides

This guide is general information, not legal advice, and reading it does not create an attorney-client relationship. Laws and policies change; check the sources and talk to a lawyer about your situation.