AI Meeting Notetakers Crossed from Helper to Legal Liability

AI meeting notetakers have earned their own reckoning. A federal class action now accuses Otter.ai of quietly joining meetings and recording private conversations without meaningful consent. A New York federal judge ruled in February that AI conversation recordings cannot be shielded by attorney-client privilege. And the New York City Bar Association’s ethics committee told lawyers in December that bringing an AI notetaker into a client meeting can waive privilege on its own.

The moment the cost first showed up was not in a courtroom. It was a virtual meeting at an Ontario hospital in September 2024, when an AI notetaker joined a discussion of seven patients and emailed the transcript to 65 people who never agreed to be part of it. A tool that bills itself as a silent helper has, in roughly two years, drawn a federal lawsuit, a binding ethics opinion, a federal court ruling on privilege, and a stack of compliance guides. Privacy lawyers across North America are starting to tell clients the same thing: the meeting bot is not yours. Professional teams that once invited the bot in by reflex are now asking whether they should send it back out.

The Quietest Participant in the Room

Otter.ai’s flagship product, Otter Notetaker, has recorded and processed more than one billion meetings, the firm says. A 22 December 2025 press release from the company put the service at more than 35 million users. The bot joins scheduled calls as a visible participant and captures audio, transcripts, screenshots, and speaker voiceprints. A federal lawsuit filed months earlier describes the system as joining meetings participants never agreed to.

Otter is not alone. Fireflies.ai, Fathom, Trint, Read AI, and Microsoft’s Copilot all offer to listen, transcribe, and summarize virtual meetings, often without anyone in the room understanding the privacy and consent obligations that come with them. Boards, doctors, and young professionals have become some of the most fervent adopters, according to the same privacy and employment lawyers tracking the rollouts.

The recordings do not stop at the discussion a human would type down. AI notetakers capture the weather talk, the hobby chatter, the political aside, and the off-the-record asides as faithfully as the agenda items, said Nicolas Joubert, a Winnipeg-based partner at the firm MLT Aikins. They don’t know what’s small, he added. The same lawyers say the tools often arrive without any participant in the room noticing.

What the tools can’t tell, Joubert and other privacy lawyers say, is when a meeting has gone in camera, the legal term for an off-the-record stretch that should not be on tape. When the agenda ends and the real conversation begins, the bot just keeps typing.

• 35 million users on Otter.ai’s transcription tools, per the company’s 22 December 2025 press release
• 1 billion+ meetings processed by Otter.ai since the company was founded in 2016
• 12 states now require all-party consent to record a private conversation, per Mayer Brown’s June 2026 compliance guide
• 4 lawsuits consolidated into one federal class action against Otter.ai in the Northern District of California
• 65 invitees received an unintended transcript of a doctor meeting about seven patients in Ontario in 2024

Seven Patients and 65 Strangers Who Saw the Notes

The Ontario hospital breach is now the canonical case. The province’s information and privacy commissioner posted a letter last year describing a virtual meeting where doctors discussed seven patients. One of the invited doctors had left the hospital in June 2023.

Because he was using his personal email for the meeting and had installed Otter.ai on his personal device, the transcription service was able to join and quietly record the call. No one in the meeting consented. Otter.ai did not ask. The transcript and a generated summary were emailed automatically to 65 people who had never asked to be part of it.

Tools like this one don’t know the difference between the meeting’s substance and the side chat, said Teresa Scassa, Canada Research Chair in information law and policy at the University of Ottawa. They also don’t know when a discussion goes in camera, so when private portions of board or court conversations happen, the tool keeps recording. That detail matters for every profession that uses these platforms.

All of a sudden, all those in-camera discussions have just been distributed to the entire meeting mailing list and that’s obviously a huge problem.

Teresa Scassa, Canada Research Chair in information law and policy at the University of Ottawa, made the point in an interview with the Canadian Press that ran nationally on 1 March 2026. Scassa and Joubert are not alone. Across the legal profession, the loudest critics of these bots are now lawyers who once used them the most.

When the Lawyers Stopped Hitting Record

For a long time, lawyers were among the heaviest users of AI notetakers. They are now their loudest critics.

The New York City Bar Association’s Formal Opinion 2025-6, issued 22 December 2025, told lawyers that bringing an AI notetaker into a client meeting without informed consent can waive attorney-client privilege on its own. On 13 February 2026, Judge Jed S. Rakoff of the Southern District of New York ruled in United States v. Heppner that AI tools are not attorneys and that conversations with them are by definition outside the privilege. The Heppner case involved a senior executive accused of securities fraud who used Anthropic’s Claude AI to analyze his own defense. The court rejected his claim that the AI records were privileged.

Outside the courtroom, ethics memos have stacked up. A June 2026 compliance guide for global AI notetakers catalogs the cross-jurisdictional minefield, including the federal Wiretap Act’s one-party consent default, twelve states with all-party consent rules, GDPR exposure in Europe, and biometric voiceprint rules that kick in when speakers are automatically identified. The guide treats AI notetaker compliance as a governance problem rather than a side project.

The Mayer Brown guide makes one practical point sharper than the rest: anything typed into a consumer AI tool should be treated as if it were posted publicly, because what looks like a private transcript is, for legal purposes, more like an email blast. A meeting recording captured by a third-party bot that stores its transcripts on a vendor’s cloud server can travel that path too. Once privilege is treated that way, the next question for lawyers is whether to let the bot listen. Most firms now answer no.

The cases privacy lawyers follow most closely are the meetings that cross borders. One Zoom call today can pull participants from three jurisdictions, two regulatory regimes, and one in-camera segment that should not be on tape.

1. September 2024: an AI notetaker joins a doctor meeting at an Ontario hospital about seven patients and emails the transcript to 65 people, per Ontario’s information and privacy commissioner.
2. 15 August 2025: Justin Brewer files a class action in U.S. District Court for the Northern District of California accusing Otter.ai of secretly recording meetings for AI training.
3. 22 December 2025: New York City Bar releases Formal Opinion 2025-6, warning lawyers that AI notetakers can waive attorney-client privilege without informed consent.
4. 13 February 2026: SDNY Judge Rakoff rules in United States v. Heppner that AI conversation recordings are not protected by attorney-client privilege.
5. June 2026: Mayer Brown publishes a global compliance guide on AI notetakers across the U.S., EU, China, and Brazil.

The Class Action That Named Otter

The consolidated federal suit against Otter.ai, In re Otter.AI Privacy Litigation (case 5:25-cv-06911), was built from four class actions filed in the Northern District of California between August and September 2025. Lead plaintiff Justin Brewer, of San Jacinto, California, says Otter’s OtterPilot service joined a private conversation and recorded it without meaningful consent. Walker, Theus, and Winston followed with three more suits within a month. On 22 October 2025, Judge Eumi K. Lee consolidated the cases and scheduled a motion-to-dismiss hearing for 20 May 2026 at the San Jose federal courthouse.

Other incidents in the public record hint at what consent-by-default looks like in the wild, according to consolidated case records and trade-press reporting. An AI researcher told The Washington Post, as cited by NPR, that Otter recorded a Zoom meeting with investors after he had left the call, then emailed him the transcript; some of the material was intimate and confidential, and a deal died. A Politico correspondent who uses Otter wrote that she does not bring the tool into interviews with Uyghur human rights activists, for fear the recordings could end up somewhere they should not. Fireflies.ai was hit with its own BIPA suit, Cruz v. Fireflies.AI Corp., in Illinois in December 2025 alleging the same voiceprint-capture mechanics. The University of Washington, Chapman University, and the University of California at Riverside have banned Read AI from their Zoom and Teams integrations, a step taken without any litigation.

The complaint names the cloud storage and the voiceprints Otter generated to label who said what. Damages under the federal Electronic Communications Privacy Act, California’s Invasion of Privacy Act, and Illinois’s Biometric Information Privacy Act run into statutory millions even before class certification. A consolidated complaint was filed in December 2025, and Judge Lee’s upcoming ruling on the motion will be the first federal test of whether decades-old wiretap statutes reach an AI bot sitting in the corner of a video call.

If they accuse us, then they could accuse everyone else, all the tools you heard about doing meeting notes. My view is that we are on the right side of history. We’re building this new AI revolution. If you want AI to help, you need to put AI in the meetings.

That was Otter.ai chief executive Sam Liang in a 7 October 2025 TechCrunch interview. The Mayer Brown guide’s framing of the legal stakes, and the Heppner reasoning, are now the counter-position every regulator and enterprise procurement team is bracing for.

What Professionals Are Asking Before They Hit Record

The response across professional firms is starting to look the same. Recruiters, in-house counsel, and HR teams now run a common set of consent and data-handling questions before they invite a bot into a meeting, and they are starting to demand answers on paper. The most important questions, according to privacy lawyer Nicolas Joubert, are where the data lives, whether the vendor sells access to it, whether the vendor trains models on the recordings, and whether any third party can reach the data. Vendors that cannot answer them clearly are starting to lose their place on the approved software list.

Red Seal Recruiting Solutions, a Victoria, B.C. hiring firm, has used the interviewing platform HoneIt for about four years. President Kael Campbell said the firm’s clients rarely object to the bot, the firm’s recruiters notify participants before any recording starts, and the firm verifies the output before sending it on. If a client objects, Campbell said, the company reverts to typing or handwriting notes. Before deploying the tool, the firm combed through the vendor’s terms of service and verified compliance with both Canadian and European data laws. Vendors do not all volunteer that level of disclosure.

Who Actually Owns the Recording

Under the federal Wiretap Act, recording a call is legal if at least one party consents. Twelve states, including California, Connecticut, Florida, Illinois, Maryland, Massachusetts, Michigan, Montana, Nevada, New Hampshire, Pennsylvania, and Washington, go further and require every participant to consent. Three states, Connecticut (already on the all-party list), Delaware, and New York, separately require employers to give written notice before recording workers on the job. A single Zoom call can pull participants from three jurisdictions, three consent regimes, and one biometric voiceprint at a time. The compound risk grows quietly.

AI notetakers complicate every layer of that test. Most platforms create de facto voiceprints of each speaker as they label who said what. Several privacy regimes, including Europe’s GDPR and Brazil’s LGPD, treat voice biometrics as sensitive data on their own. A bot that works in New York might be illegal in California for a recorded segment, but legal for a live one.

Privacy lawyers and bar opinions now frame the consent question more strictly than the federal default allows. Joubert’s plain-language version from a Canadian Press interview: you probably wouldn’t want to stand on a busy street corner with a sandwich board with that information during rush hour; it’s really no different. That’s the test many firms now apply to recording bots. Professionals that take that test seriously extend it to any bot they cannot fully audit on the call.

Frequently Asked Questions

Why are lawyers warning against AI meeting notetakers?

Lawyers warn because the December 2025 New York City Bar Formal Opinion 2025-6 told them that bringing an AI notetaker into a client meeting without informed consent can waive attorney-client privilege on its own. A 13 February 2026 ruling in United States v. Heppner went further, holding that AI conversation recordings are not protected by attorney-client privilege at all. Mayer Brown’s June 2026 compliance guide treats the bots as a governance problem that has to be managed before the meeting, not after.

What does the Otter.ai class action actually say?

The consolidated federal suit, In re Otter.AI Privacy Litigation (case 5:25-cv-06911), was built from four class actions filed in the Northern District of California between August and September 2025. Lead plaintiff Justin Brewer is from San Jacinto, California. The plaintiffs allege Otter’s OtterPilot service joined meetings without the consent of non-account-holder participants and that Otter used the captured audio and voiceprints to train its AI. A consolidated complaint was filed in December 2025, and Judge Eumi K. Lee’s motion-to-dismiss hearing was set for 20 May 2026 at the San Jose federal courthouse.

What are the consent rules for AI meeting tools?

In the United States, the federal Wiretap Act allows recording when at least one party to a conversation consents. Twelve states require every participant to consent. Three states, Connecticut, Delaware, and New York, require employers to give written notice before recording workers on the job. Outside the US, the GDPR and Brazil’s LGPD treat voice biometrics as sensitive and impose additional notice and retention rules.

What steps keep an AI notetaker from recording a meeting?

Most meeting platforms let hosts block bots in settings, decline bot invites, or set waiting-room rules that filter out automated participants by default. Privacy lawyers also recommend a written notice to all attendees before recording, a written vendor risk assessment, and a clear retention rule for any recording that does happen. The University of Washington, Chapman University, and the University of California at Riverside have all banned Read AI from their Zoom and Microsoft Teams integrations. Several law firms now turn the bot off by default for legal, HR, and student-record meetings.

Disclaimer: This article covers legal and workplace-privacy considerations around AI meeting notetakers. It is informational and not legal advice. Recording and consent rules vary by jurisdiction; consult a qualified attorney in your jurisdiction before deploying these tools in any meeting that handles sensitive, regulated, or privileged information. Figures and case statuses are accurate as of publication.