Your Client Used AI to Prepare for Your Meeting — And Now It's Not Privileged
Published March 13, 2026 · By The Crossing Report · 8 min read
Summary
On February 17, 2026, a federal judge in the Southern District of New York issued the first ruling of its kind: documents a client creates independently using AI tools are not protected by attorney-client privilege. Multiple major law firms have already issued client alerts. Here is what your small firm needs to know, and three things to do this week.
What the Court Actually Decided
The case is United States v. Heppner. The defendant, Bradley Heppner, was facing federal charges related to securities and wire fraud. After his arrest in November 2025, agents seized electronic devices containing 31 documents he had created using Anthropic's Claude chatbot — materials he said he had prepared to facilitate discussions with his attorneys.
When prosecutors sought those documents, Heppner argued they were protected by attorney-client privilege. Judge Jed Rakoff disagreed. The AI-generated materials were not privileged. Work product protection also failed.
The core of Judge Rakoff's ruling comes down to one principle: attorney-client privilege requires confidentiality, and sharing information with a third-party AI platform breaks it.
When a client independently uses a consumer AI tool — without attorney direction, without attorney involvement — to generate documents, the client has voluntarily shared their information with a third party. That act of sharing, the court found, waives the confidentiality requirement that privilege depends on.
The court's specific language is worth understanding: “AI is not legal counsel.” A consumer AI chatbot owes users no duties of loyalty or confidentiality. The platform retains the right to use that data for training and other purposes. By using it, a client has made an independent choice to involve a third party in their legal matter — a choice the attorney did not direct and may not have known about.
The work product doctrine failed for related reasons. Work product protection covers materials prepared by or for an attorney in anticipation of litigation. The 31 documents Heppner created were prepared by him, independently, using a consumer tool. The attorney wasn't directing the work; the attorney wasn't even involved yet.
The practical implication
If your client opened ChatGPT, Claude, or any other consumer AI tool, used it to generate notes, talking points, or analysis about their legal matter, and then brought those materials to a meeting with you — those materials are not protected. A court can order them produced.
The Kovel Exception — The Path to Privilege That Still Works
The Heppner ruling does not say all AI use in legal matters destroys privilege. It says client-directed, independent AI use does. And there's an important distinction the court preserved.
The Kovel doctrine, established in a 1961 Second Circuit case, allows third parties to participate in attorney-client communications when the attorney directs their involvement to facilitate legal advice. The classic example: an attorney hires an accountant to translate complex financial information so the attorney can advise the client. The accountant's involvement doesn't break privilege because the attorney is directing the engagement.
The Heppner ruling acknowledges that this framework applies in the AI context. Attorney-directed AI use — where the attorney controls the tool, directs how it's used, and deploys it as part of rendering legal advice — has a much stronger privilege argument than client-initiated use.
For your practice, this means the relevant distinction is:
- •Client-directed AI (not privileged): Client uses ChatGPT or Claude to prepare questions for you, draft a timeline of events, research their situation, or organize documents before a meeting. The client did this independently, without your direction.
- •Attorney-directed AI (potentially privileged): You use AI to analyze client documents, draft communications, research case law, prepare strategy memos. You are directing the tool as part of your legal work. The Kovel analysis may apply.
This distinction has immediate implications for how you communicate with clients and how you document your own AI use.
Three Things to Do This Week
1. Send a Client Advisory
Your active clients need to know about this ruling. The practical risk is real: a client who uses AI to prepare for your meetings, then discovers in litigation that those materials are discoverable, has a legitimate question about whether they were warned. An advisory email creates a clear record that they were.
The message doesn't need to be formal or lengthy. Three sentences is enough:
“A recent federal court ruling (United States v. Heppner, SDNY, February 2026) established that documents a client creates independently using AI tools to prepare for meetings with their attorney are not protected by attorney-client privilege. If you use AI in connection with your legal matter — to draft questions, prepare documents, research your situation, or organize information — please contact us before doing so. We want to make sure your communications remain fully protected.”
Send this to active clients today. It takes 20 minutes and creates a documented record. If your clients have matters in active litigation or potential litigation, this is urgent.
Premium Content
Actions 2 & 3: Engagement Letter Update and Internal AI Protocol
The complete playbook for updating your engagement letter and building an internal AI governance protocol — the two steps that protect your firm and every future client relationship. Premium subscribers also get every future issue in full.
- 2.Update Your Engagement Letter — the one paragraph addition that sets AI governance expectations from day one.
- 3.Build an Internal Protocol for AI Use — the documentation practice that distinguishes your supervised use from the independent client use at issue in Heppner.
Free weekly digest. No spam. Unsubscribe anytime.
$19/month · Cancel anytime · First issue free
Who Else Needs to Know About This
The immediate audience for the Heppner ruling is attorneys. But the underlying principle — that independently sharing confidential information with a third-party AI platform breaks privilege — extends to other professional contexts.
Accountants working on matters with any litigation exposure should be aware that the same analysis may apply to any privilege claim they or their clients might assert. Consultants working on matters involving regulatory investigations or potential litigation face a similar question. The specifics will differ by jurisdiction and professional context, but the core principle is the same: the AI tool is a third party, and using it independently, without professional direction, may have privilege consequences.
If your clients include attorneys or other regulated professionals, a note about this ruling in your own client communications is worth considering.
Frequently Asked Questions
What did the Heppner ruling decide about AI and attorney-client privilege?
In United States v. Heppner (SDNY, February 2026), Judge Jed Rakoff ruled that documents a client creates independently using AI tools are not protected by attorney-client privilege. The court held that using a consumer AI platform constitutes sharing information with a third party, which waives the confidentiality requirement that privilege depends on. The ruling also denied work product protection because the AI-generated materials were prepared by the client independently, without attorney direction.
Does using ChatGPT or Claude break attorney-client privilege?
It depends on who is using it and under whose direction. Under the Heppner ruling, a client who independently uses a consumer AI tool to prepare documents for attorney meetings loses privilege protection on those materials. However, attorney-directed AI use — where the attorney controls the tool as part of rendering legal advice — has a stronger privilege argument under the Kovel doctrine.
What is the Kovel exception and does it apply to AI?
The Kovel doctrine (established in a 1961 Second Circuit case) allows third parties to participate in attorney-client communications when the attorney directs their involvement to facilitate legal advice. The Heppner ruling acknowledged this framework applies to AI: attorney-directed AI use has a much stronger privilege argument than client-initiated use.
What should small law firms do after the Heppner ruling?
Three immediate steps: (1) Send a client advisory email informing active clients that independently using AI to prepare for attorney meetings may not be privileged. (2) Update your engagement letter to include AI governance expectations. (3) Build an internal protocol for attorney-directed AI use that documents supervision and distinguishes your use from the independent client use at issue in Heppner.
The Bigger Picture
The Heppner ruling is the first of its kind, but it will not be the last. Courts are writing the rules of AI-era privilege in real time, and they're writing them based on the factual records in cases like this one.
The good news for small firm owners: the attorney-directed AI use your firm does is on the right side of this analysis. You're using AI as a professional tool, under your supervision, to deliver better legal services. That's exactly the fact pattern the Kovel doctrine was designed to protect.
What this ruling really asks of you is to be explicit about that. Document the supervision. Update the engagement letter. Tell your clients what they should and shouldn't do. Three steps, none of which take more than an hour, and all of which position your firm as the attorney who gets ahead of issues rather than the one who reacts to them.
The attorneys who are best positioned in the AI era are the ones who are proactive. This is a proactive move.
Your Action This Week
Draft the client advisory email. Use the three-sentence template above, adapt it to your voice, and send it to active clients today.
If you have clients currently in litigation or with matters where privilege might matter, send this today. If your practice is primarily transactional or advisory, send it this week.
This is the rare compliance task that's also a client relationship opportunity. You're the attorney who saw a federal ruling, understood what it meant for your clients, and told them about it before they discovered it the hard way. That's exactly the kind of value that justifies a long-term client relationship — and that no AI tool can replicate.