Your AI Notetaker Might Be Waiving Privilege: What United States v. Heppner Means for Small Firms
A federal judge ruled a client's chats with a consumer AI tool weren't privileged. Here's what that means for AI notetakers in your client calls, and a five-step fix.
A federal judge just decided that talking to a public AI tool is not the same as talking to your lawyer.
In February 2026, a federal judge in Manhattan ruled on a question nobody in your firm wanted answered. Are a client's chats with a consumer AI tool protected by attorney-client privilege?
The answer, in that case, was no.
Most small firms have not thought about what that means for the AI tools already running in their own meetings. This post covers the ruling, what it does and doesn't decide, and a fix you can finish this week.
What the court actually decided
United States v. Heppner is a criminal case in the Southern District of New York. The defendant, Bradley Heppner, used the consumer version of Claude on his own. He fed it information he had learned from his lawyers. He asked it for reports on defense strategy. Then he shared the results with counsel. (Proskauer summary)
Judge Jed Rakoff ruled from the bench on February 10, 2026. He issued a written opinion on February 17. (Debevoise)
He held two things:
- No attorney-client privilege. The chats were not confidential. The tool's privacy policy tells users the company collects data and may disclose it to third parties. So the court found Heppner had no reasonable expectation of confidentiality.
- No work-product protection. Heppner acted on his own, not at his lawyers' direction.
One law firm called it the first ruling of its kind. (O'Melveny)
What it does not decide
Be precise here. Your clients will ask.
The ruling did not say that using AI waives privilege by default. It involved a consumer tool, a client acting alone, and a specific privacy policy. Open questions include whether a lawyer's own AI use would be protected, and whether a client using AI at counsel's direction could claim protection. (Proskauer)
So this is not a ban on AI. It is a warning about one thing: the vendor's terms matter to your privilege.
Why notetakers belong in the conversation
The case involved a chatbot, not a meeting recorder. Applying it to notetakers is our reading, not a holding.
But the logic is easy to follow. If a court looks at a tool's terms to decide whether a conversation was confidential, then any AI tool that hears a client call invites the same question. Where does the transcript go? Who can read it? Does the vendor train on it?
The ethics guidance already points this way. ABA Formal Opinion 512, issued July 29, 2024, says lawyers must weigh the risk that client information will be disclosed to people outside the firm before they put it into a generative AI tool. It also says informed client consent is needed before entering client information into a self-learning tool. (ABA Business Law Today)
Heppner is the first case we've seen where a court tested that kind of confidentiality question against a real privilege claim.
A five-step fix for this week
None of this needs new software. It needs a few hours and a decision.
1. List every AI tool that touches client information. Check video platforms, phone systems, transcription apps, and browser add-ons. Ask staff what they use. You will find tools you didn't approve.
2. Read each vendor's terms on three points. Does it keep your data? Does it train models on it? Can it disclose it to third parties? If you can't find the answer, treat it as a no.
3. Set a notetaker rule. No AI recorder joins a client meeting unless the firm approved it and the client knows. Make the host responsible for checking the participant list before the call starts.
4. Add AI language to your engagement letter. Cover two things. How the firm uses AI tools, and what you ask of the client. Ask clients not to paste your advice into public AI tools.
5. Give staff a 30-second script. "We use an approved tool to take notes. It doesn't share your information. If you'd rather we didn't, tell us and we'll take notes by hand."
Then write the date and the tool list in a file. If anyone ever asks what you did, you have an answer.
Talking to clients about it
The most useful instruction may be the simplest: don't run our advice through a public chatbot.
Clients mean well. They want to understand their case. They don't know that a chat with a consumer tool may not be confidential. Tell them at intake, and again when the case gets stressful.
The bottom line
Heppner is one trial-court ruling, and courts will keep testing this. Treat it as a reason to know your tools, not a reason to panic.
The firms that come out fine will be the ones that can say, in one paragraph, which tools they use, what those tools' terms say, and what clients were told.
That is a small governance job, and it's doable now. For related reading, see our post on state AI regulation and lawyers.
Get a Free Audit
Not sure what your firm's AI tools do with client data? We'll review your setup and give you a plain-English list of gaps and fixes.