Your AI Chatbot Is Not Your Lawyer

June 4, 2026 | Comments Off on Your AI Chatbot Is Not Your Lawyer
Posted by Kurt E. Anderson

Ever been tempted to type a sensitive legal question into ChatGPT, Claude, or Google Gemini? You’re not alone. But here’s the catch: that AI chatbot you’re confiding in? It’s not your lawyer. And as a New York federal court just made clear, your conversations with it aren’t privileged — meaning your “confidential” information may already be fair game.

The Heppner Case: A Cautionary Tale

In United States v. Heppner (S.D.N.Y. Feb. 2026), the defendant was indicted on criminal charges. After receiving a grand jury subpoena and learning he was a federal target, he did what many people might do today — he turned to AI. Using Anthropic’s Claude, he generated roughly thirty-one reports outlining potential defense strategies.

The problem? Heppner did this entirely on his own, without any direction from his attorneys. He later shared the AI-generated reports with his lawyers. When FBI agents seized the documents during a home search, his defense team cried privilege. The government disagreed — and so did the judge.

Why the Privilege Claim Failed

Attorney-client privilege requires three things: (1) a communication between client and attorney, (2) intended to be confidential, and (3) made for obtaining legal advice. Heppner’s AI chats struck out on at least two, arguably all three.

Claude isn’t a lawyer. As Judge Rakoff put it, recognized privileges require “a trusting human relationship” with “a licensed professional who owes fiduciary duties and is subject to discipline.” No such relationship exists with an AI.

The communications weren’t confidential. Claude’s privacy policy lets Anthropic collect user inputs and outputs, use that data to train the model, and disclose it to third parties — including the government. By typing his defense strategy into Claude, Heppner was essentially cc’ing a third party that reserved the right to forward everything to federal prosecutors.

No legal advice was being sought (from a lawyer). Claude disclaims providing legal advice, and Heppner’s attorneys hadn’t directed him to use the platform. Judge Rakoff noted that attorney-directed AI use might yield a different outcome — but that escape hatch wasn’t available here.

The court also rejected the work product argument. The documents weren’t prepared by or at counsel’s direction. Sharing them with lawyers after the fact did not, as the court put it, “alchemically change” unprivileged materials into privileged ones. (Great word choice, Judge Rakoff.)

The Confidentiality Problem Is Bigger Than Privilege

The privilege issue grabs headlines, but the underlying confidentiality risk is arguably worse. As Ogletree Deakins observed, when you input information into a public AI tool, that data travels to remote servers where it may be stored, reviewed by employees, used to train future models, or disclosed to third parties. ChatGPT, Claude, and Gemini all have terms of use permitting some form of data collection.

Translation: this isn’t just about privilege waiver. If you input trade secrets, proprietary strategies, personnel matters, or other sensitive information into consumer AI platforms, you risk losing confidentiality permanently. Once data is ingested by a platform that trains on user inputs, there’s no “undo” button. As one law firm advisory put it bluntly: anything you submit “may not remain confidential.”

A Quick Counterpoint: Warner v. Gilbarco

One week before Heppner, the Eastern District of Michigan reached a different result in Warner v. Gilbarco. There, a pro se plaintiff’s use of ChatGPT was found to be protected work product. The distinction: Warner was representing herself, effectively acting as her own counsel, and work product protection is waived only by disclosure to an adversary — not to an AI. Interesting? Yes. But if you have a lawyer, expect Heppner to be the more influential decision.

The Bottom Line

The moral is simple: don’t use consumer AI platforms to research legal matters unless your attorney specifically directs you to. Beyond that:

Read the fine print. If a platform’s terms permit data collection or third-party disclosure (and most do), nothing you share is confidential.

Talk to your lawyer first. If you’re in litigation, under investigation, or anticipating a legal dispute, consult counsel before using any AI tool in connection with the matter.

Sharing with AI = sharing with a third party. You wouldn’t cc a stranger on a privileged email to your attorney. Typing that info into a chatbot is, legally speaking, the same thing.

If you’ve already used AI, stop digging. Don’t compound the problem by emailing AI outputs around. Tell your attorney immediately. Remember: sharing AI-generated content with your lawyer after the fact doesn’t retroactively make it privileged.

Think beyond privilege. Even outside litigation, inputting trade secrets or sensitive business information into consumer AI tools risks permanent loss of confidentiality. Once it’s in, it’s not coming out.

Heppner is just one district court decision, and the law here is still evolving. But the takeaway is clear: the convenience of AI doesn’t come without legal risk. Your chatbot is many things — but it’s no substitute for your lawyer.