AI note takers now sit in many client calls. No court has found a waiver from one yet, but the privilege rulings and ethics guidance point to real risk.
They can, and the risk depends mostly on the tool’s terms and on who turned it on. As of this writing, we found no reported court decision holding that an AI meeting notetaker by itself waived attorney-client privilege. But the reasoning courts have used for other AI tools, and the ethics guidance written since, give lawyers good reason to treat these bots as a third party in the room until they have checked the contract.
Searches for “AI note taker attorney client privilege” now show up as Google autocomplete suggestions, which tracks with how fast these tools spread through Zoom, Teams and Google Meet calls.
Why Is a Notetaker Different From a Paralegal Taking Notes?
Privilege generally requires a communication made in confidence. A human paralegal working for the firm is inside the circle of confidentiality. A notetaker run by an outside vendor may not be, depending on what the vendor can do with the recording and transcript.
That is the logic Judge Jed Rakoff applied in United States v. Heppner in February 2026, a case about a defendant’s chats with a consumer AI tool rather than a notetaker. The court found no reasonable expectation of confidentiality because the provider’s privacy policy allowed it to use and disclose inputs. Lucian Pera’s September 2026 ethics column in ABA Law Practice magazine applies the same test to notetakers directly. Asked about assistants used on Zoom or Microsoft Teams, he answers: “Same answers. Read their terms and check what exactly the vendor retains and where.”
What Are the Specific Risks?
The first is confidentiality. A vendor that stores transcripts, uses them to train models, or can share them under its policy weakens the case that the conversation stayed confidential.
The second is discoverability. Blank Rome’s guidance on AI notetakers points out that these tools create “vast amounts of potentially discoverable business records,” capturing candid conversations that were never written down before. A transcript that exists can be requested, and it has to be preserved once litigation is reasonably anticipated. Pera’s column says a client’s AI prompts must be preserved when litigation is imminent, and lawyers are very likely under the same duty.
The third is the client’s own bot. If a client runs a notetaker on a call with counsel without the lawyer directing it, the Heppner reasoning on work product suggests the output may get no protection. Blank Rome recommends advising clients not to use their own AI notetakers on calls about active litigation.
The fourth is accuracy. A transcript summary that misstates what a lawyer said can travel further than the original conversation. Ethics rules on competence still apply to reviewing AI output.
What Are Firms Doing About It?
The practical guidance from Blank Rome and Pera points to a few steps. Approve specific notetaking tools after reviewing the vendor’s terms on training, retention, deletion and access. Get client consent before AI tools record a session; Blank Rome notes this is often handled in the engagement letter. Limit what kinds of discussions can be recorded at all. And treat transcripts as records subject to retention schedules and legal holds.
The American Bar Association’s GPSolo eReport covered these confidentiality risks in its September 2025 issue, before the 2026 privilege rulings. In December 2025 the New York City Bar Association issued Formal Opinion 2025-6 on AI tools that record and summarize client conversations; as Blank Rome summarizes it, lawyers should get a client’s consent before recording and should review every transcript and summary for accuracy. For a broader checklist of vendor questions, see our guide on legal AI and client confidentiality.
What to Watch
The first ruling squarely on a notetaker transcript is likely to come out of a discovery fight, not a malpractice case. Watch protective orders, too. The court in Morgan v. V2X added language limiting uploads to AI platforms that train on inputs, and similar clauses could easily reach meeting recordings.
Partly Real. No court has yet found a waiver from a notetaker alone. The underlying risk, an outside vendor holding a transcript of privileged advice, is real enough that firms are writing rules now.
Common Questions
Does an AI note taker waive attorney-client privilege?
No reported decision has held that yet. Courts have looked at whether a provider’s terms allow it to use or disclose data, which is the same question a notetaker raises.
Are AI meeting transcripts discoverable?
They can be. Commentators note that transcripts are records like any other and may need to be preserved once litigation is anticipated.
What should a firm check before approving a notetaker?
Whether the vendor trains on recordings, how long it keeps them, who can access them, whether they can be deleted, and whether the contract includes confidentiality obligations.
Should clients run their own notetakers on calls with counsel?
Several firms advise against it on litigation matters, since output a client creates on their own may not get work product protection.
In this story
See every round in the Funding Tracker