How Your AI Chats Can Be Used Against You
A recent federal court ruling sends a clear warning to anyone who has discussed legal matters with an AI tool like ChatGPT or Claude: your chats are potentially discoverable by a court, a government agency, or an opposing party in a lawsuit.
I. A Case That Made Legal History
In February 2026, a federal judge in New York issued what may be the first ruling of its kind in the entire country. In United States v. Heppner, the court decided that private conversations a criminal defendant had with an AI platform were not protected—meaning that the government could read every word.[1]
Bradley Heppner was charged with securities fraud and related crimes.[2] While under investigation, and without being asked to do so by his lawyers, he used the AI platform Claude to draft what he called “defense strategy” documents.[3] He later shared those documents with his attorneys and argued they should be kept private (protected by the attorney-client privilege and the work product doctrine).[4]
The court disagreed. It ordered the documents turned over to the government.[5]
II. Why the Court Said No
The court found that the AI documents failed to qualify for protection for three main reasons:
First, Claude is not a lawyer. (Neither is ChatGPT or any other AI platform.)
The attorney-client privilege only protects communications between a client and an actual attorney.[6] Because Claude is an AI and not a licensed professional, the privilege simply does not apply. In fact, when the government asked Claude whether it could give legal advice, Claude itself answered: “I’m not a lawyer and can’t provide formal legal advice or recommendations.”[7]
Second, AI chat conversations are not confidential. When you use a platform like Claude, you agree to a privacy policy. That policy allows the company to collect your inputs and outputs, use them for training the AI, and share them with third parties (including government regulators).[8] The court found that Heppner had no reasonable expectation of privacy once he typed his thoughts into a third-party platform.[9]
Third, the documents are not protected work product. The work product doctrine shields materials that a lawyer (or the lawyer’s agent) prepares in anticipation of a lawsuit.[10] Because Heppner acted entirely on his own, and his lawyers never directed him to use Claude, the documents did not reflect the mental strategy of his legal team. They were simply his personal notes, dressed up in AI output.[11]
III. What This Means for You
This case is a warning for everyone, not just people facing criminal charges. Here is the core lesson: anything you type into a public AI platform is potentially discoverable by a court, a government agency, or an opposing party in a lawsuit.
When you ask an AI tool about a potential legal situation, those conversations may become available to the other side of that dispute. This danger is only compounded by the risk of hackers gaining access to your chats—unsurprisingly, they will find it far easier to access your data when it is stored in the free versions of ChatGPT or Claude than when it is securely locked away in a law firm’s files.
The temptation to use AI for legal thinking is understandable. These tools are fast, available around the clock, and appear knowledgeable. But as the Heppner court made clear, using AI is nothing like talking to your attorney. There is no privilege. There is no confidentiality guarantee. And there is no fiduciary duty owed to you.
Using AI tools for legal advice does not replace competent and qualified legal counsel.
The judge summed it up well: AI’s novelty does not place it beyond longstanding legal rules. The law has not changed just because the technology has.
IV. Practical Steps to Protect Yourself
If you are facing a legal matter (or even suspect that one may be on the horizon) keep these points in mind.
-
Do not use public AI tools to plan your legal strategy. Even if you intend to share the results with your lawyer later, a non-privileged document does not become privileged simply because you handed it to an attorney. The court in Heppner was clear on this point.
-
Do not assume that your AI conversations are private. Review the privacy policy of any platform you use. Most major AI providers reserve the right to store, use, and share your data in ways that could later be disclosed.
-
If your attorney asks you to use a specific tool or process to gather information, follow those instructions carefully and document that you were directed to do so. The Heppner court noted that the outcome might have been different had Heppner’s lawyers told him to use Claude, because then it might have been treated as part of the attorney’s own work process.[12]
-
Retain qualified legal counsel. A conversation with your attorney is protected. A conversation with a chatbot is not.
Summary
The Heppner decision is a landmark moment in the law’s effort to catch up with technology, and it will not be the last ruling of its kind. Courts are watching how people use AI, and they are applying old rules to new tools.
The ideal approach is simple: treat your AI conversations the way you would treat a conversation in a crowded coffee shop. Do not say anything you would not want a stranger to hear. Then call a qualified attorney for the discussions that actually need to stay private.
Our firm helps individuals and businesses understand their legal rights and navigate complex disputes. If you have questions about privilege, legal strategy, or how to protect your communications, contact Simms Showers today for a confidential consultation. The conversation you have with us—unlike the one you have with an AI—stays private.
Author, Micah Schachtner, Esq., practices in the areas of intellectual property, business law, and church/nonprofit law.
[1] United States v. Heppner, No. 25 Cr. 503 (JSR), 2026 WL 436479, at *1–2 (S.D.N.Y. Feb. 17, 2026). The court noted it was “unaware of, and the parties have not identified, any case to date that has presented this issue.” Id. at *1 n.1.
[2] Id. at *1.
[3]Id.
[4] Id.
[5] Id. at *3–4.
[6] Id. at *2 (quoting United States v. Mejia, 655 F.3d 126, 132 (2d Cir. 2011)).
[7] Id. at *3 (quoting ECF No. 23-6 at 1–2).
[8] Id. at *2–3.
[9] Id. at *3 (citing Mejia, 655 F.3d at 132–34; In re OpenAI, Inc., Copyright Infringement Litig., No. 25 MD 3143, ECF No. 1021 at 3 (S.D.N.Y. Jan. 5, 2026)).
[10] Id. at *3–4 (quoting In re Grand Jury Subpoenas Dated Mar. 19, 2002, & Aug. 2, 2002, 318 F.3d 379, 383 (2d Cir. 2003); United States v. Nobles, 422 U.S. 225, 238 (1975)).
[11] Id. at *4 (citing ECF No. 23-5).
[12] Id.
