A Federal Judge Just Ruled Your Chatbot Conversations Aren't Privileged. Here's What That Means.

Here is a sentence worth reading twice: a federal judge has ruled that a defendant's conversations with an AI chatbot — questions he asked about his own criminal case — were protected by neither attorney-client privilege nor the work-product doctrine. The government got them.

If you have ever typed the details of a workplace dispute, a termination, or a brewing lawsuit into a consumer chatbot, this decision is about you.

What happened in Heppner

In United States v. Heppner, No. 25 CR 503 (JSR) (S.D.N.Y.), Judge Jed S. Rakoff ruled from the bench on February 10, 2026, and issued a written opinion on February 17, 2026, addressing what the court called a nationwide issue of first impression: does any privilege protect a litigant's conversations with a consumer AI product — there, Anthropic's Claude?

The court's answer was no, for three reasons:

• An AI is not a lawyer. Attorney-client privilege protects communications between a client and an attorney. A chatbot cannot form an attorney-client relationship, so there was no privilege to begin with.
• No reasonable expectation of confidentiality. The consumer platform's own privacy policy said that inputs and outputs are collected, may be used to train models, and can be disclosed to third parties — including government regulators. You cannot claim a confidential communication when the terms you clicked through say the opposite.
• No work product. The chats were not prepared at the direction of counsel, so the work-product doctrine did not apply either.

The result: the defendant's own late-night questions about his case became evidence in the government's hands.

The Texas counterpoint — and why the split is the story

Four months later, a Texas court reached a more chatbot-friendly result. In Tate Group Automotive, LLC v. Legacy Automotive Capital, LLC, Cause No. 25-BC11B-0020 (Tex. Bus. Ct., 11th Div.), Judge Grant Dorfman held on June 3, 2026, that a non-lawyer litigant's ChatGPT conversations, prepared in anticipation of litigation, can qualify as work product under Texas rules.

Before anyone celebrates: even the winning litigant in Tate Group still had to disclose which AI tool was used and all of the discovery materials fed into it. And the court flagged a trap that should make every litigator wince — feeding materials covered by a protective order into an AI may itself violate the protective order.

So the current state of the law is: one federal court says consumer AI chats get no protection at all; one Texas court says they might get work-product protection, with strings attached. Courts are split, the doctrine is moving, and nobody can tell you today, with confidence, how the judge in your case will rule.

Which leads to the only safe operating assumption.

Assume it will be read aloud in court

Discovery exists to surface what the parties knew, when they knew it, and what they did about it. A chat log in which you describe your dispute in your own words — candidly, emotionally, at 1 a.m. — is exactly the kind of record opposing counsel dreams about. After Heppner, they have a roadmap for demanding it.

Don't tell a chatbot about your lawsuit. Not because AI isn't useful — it is — but because a consumer chatbot is, legally speaking, a stranger. A helpful, articulate stranger whose terms of service say it may repeat what you tell it.

What we will tell our own clients after Heppner

• Treat consumer chatbots like a crowded elevator. Say nothing about your dispute there that you wouldn't want opposing counsel to hear.
• Don't paste case documents into consumer AI tools — especially anything produced in discovery or covered by a protective order.
• If you already have, don't delete anything; tell your attorney what tool you used and what you shared. Deleting chats once litigation is foreseeable can create a spoliation problem worse than the chats themselves.
• Keep your case-building inside privileged channels — communications with your attorney, and platforms operating under an attorney-client engagement.
• Ask any legal-tech product one question: who is on the other end of this conversation — a law firm engaged with me, or a software company's privacy policy?

Where Cloud Counsel fits — and what we won't promise

This issue is close to home for us, because conversations with Wigmore AI on Cloud Counsel are built on a fundamentally different footing than a consumer chatbot.

When you join Cloud Counsel, you sign a platform access agreement with J. Madison PLC — a law firm. Your case file, your timeline, and your conversations with Wigmore happen inside that attorney-client relationship, under a law firm's roof. The AI providers that power Wigmore are engaged as confidential agents of the firm, for the specific purpose of preserving privilege against waiver — the same way firms have long used interpreters, investigators, and litigation-support vendors. That is a very different confidentiality posture than a consumer product whose privacy policy reserves the right to train on your inputs and share them with third parties.

Now the lawyerly fine print, which we mean every word of: no one can guarantee a privilege or work-product outcome. These protections always depend on the jurisdiction, the judge, and the specific facts — and as Heppner and Tate Group show, courts are actively split on how AI fits into the doctrine. Our confidentiality framework applies to covered matters within the firm's practice areas and jurisdictions — we are a Washington, DC-area employment-law firm handling covered matters in the jurisdictions where we practice. What we can say is this: a structured case file built inside an attorney-client platform starts from a defensible confidentiality posture. A consumer chat log starts from a privacy policy that was written to protect the software company.

The bottom line

The law on AI and privilege is being written right now, one motion at a time. Until it settles, the rule of thumb is simple: if you wouldn't say it to the other side's lawyer, don't type it into a consumer chatbot. And if you want AI working on your case, put it to work somewhere the attorney-client relationship got there first.

If you're dealing with a workplace dispute and want to build your case in a confidential, structured case file, you can start a case file or contact us.

This post is attorney advertising and general information, not legal advice. Reading it does not create an attorney-client relationship, and privilege and work-product outcomes are never guaranteed — they depend on your jurisdiction and the facts of your matter. If you have questions about protecting your own communications, speak with a lawyer about your specific situation.