AI meeting notetakers went from a curiosity to a default in a few years. In 2025 and 2026 they also became the subject of class action lawsuits. If your team records calls with one, or if you are on calls that someone else's notetaker is recording, it is worth knowing what the cases say, because they turn on a question every team can answer in advance: did everyone on the call know, and agree?
This article summarises the public reporting on the main cases, keeps what is alleged apart from what a court has decided, and ends with a checklist you can use this week. It is not legal advice, and Notepik is not a law firm. Recording rules differ by country and, in the United States, by state, so take your own situation to a lawyer.
The best-known case is In re Otter.AI Privacy Litigation, No. 5:25-cv-06911, in the US District Court for the Northern District of California. It combines four proposed class actions filed between August and September 2025, and the court consolidated them in October 2025. UC Today covers the history in detail.
The plaintiffs allege that Otter's notetaker joined calendar meetings as a participant, captured audio and transcripts, did not obtain consent from the other attendees, and used captured data to train its speech recognition models. Some plaintiffs say they were not Otter customers at all. The claims rest on federal and California privacy and wiretap laws, an Illinois biometric privacy law and several common law theories.
Otter's position, as reported, is that its terms of service tell account holders to make sure they have the necessary permissions before using the notetaker, and that nothing was intercepted. The plaintiffs answer that a clause in the customer's contract does not bind the other people on the call, which matters most in states that require every participant to consent.
A separate proposed class action against Granola was filed in federal court in California in the summer of 2026. As described by the National Law Review, it alleges that Granola's software recorded a participant without telling them an AI notetaker was present and without asking permission, and that meeting content was used by default for commercial purposes, including training AI systems, unless the Granola user switched that off. The complaint contrasts this with tools that visibly join a call, announce themselves or let participants remove them.
The claims cite the federal Electronic Communications Privacy Act, the California Invasion of Privacy Act, other California statutes and common law invasion of privacy.
UC Today also reports a biometric privacy suit against Fireflies.ai in Illinois, filed in December 2025, with allegations about how voiceprints are captured. Bloomberg Law has covered the wider set of risks, including wiretapping, discovery and professional ethics.
Everything above is an allegation. The reporting we reviewed describes consolidation, filings and pending motions. It does not describe any court finding that an AI notetaker broke the law. Accounts of how far each case has progressed differ, so this article describes the claims and does not predict outcomes. For the current position, check the court dockets or a legal update from a firm you trust.
Three ideas run through the complaints. They apply to any notetaker, including ours.
The tool is a third party. A colleague taking notes is a participant. A notetaker run by a vendor is software owned by somebody else, and one theory in these cases is that this makes the vendor a separate party that has to justify its own recording, not only the customer who invited it.
Still writing meeting notes by hand?
Notepik joins the call, writes the summary, and hands you the action items before you have closed the tab.
Read our security overviewConsent belongs to everyone on the call. The person who installs a notetaker agrees to its terms. The other people on the call did not. Where the law requires every party to consent, the customer's agreement may not cover them.
What happens to the content matters as much as the recording. Both cases raise training on meeting content. A team that has settled consent can still have a second problem if the vendor keeps and reuses what it recorded.
There is also a quieter issue. Transcripts and summaries are records, and records can be requested in a dispute. Who holds them, how long they are kept and who can delete them are questions worth answering before you need the answers.
| Question | Why it matters | How Notepik answers |
|---|---|---|
| Does it announce itself to everyone on the call? | Notice is the first thing the complaints focus on | The bot joins as a named participant, Notepik Notetaker, and is always visible in the participant list. The Chrome extension does not announce itself, so it asks you to confirm that you will tell everyone before it will record |
| Who is responsible for consent? | Terms often place it on the customer | The person recording. Our consent page says so, and the extension makes you confirm it |
| Is meeting content used to train models? | Raised in both cases | No. Meeting content is not used to train models, and our processing providers are contractually barred from doing so |
| Can participants object or remove the tool? | A visible tool can be refused | The bot is a participant, so the host can remove it like any other |
| Who can see transcripts and share them? | Wider access widens the exposure | Four workspace roles. Share links are off until an owner or admin turns one on, and can be revoked at any moment |
| Can you delete a meeting? | Retention and discovery | Owners and admins can delete a meeting, and closing a workspace deletes everything after 30 days |
We would rather be plain about this than reassuring. A visible bot makes notice possible, but it does not by itself satisfy a consent law. The extension records without a bot, which is the kind of recording the Granola complaint objects to when nobody is told, so it will not start until the person recording confirms they will tell everyone, and the responsibility for doing so stays with them.
Notepik does not use meeting content to train models. Each workspace is isolated at the database level. We do not have SOC 2 or ISO 27001 certification today, and if your procurement requires one, we are not the right fit yet. Our security page lists what we do and do not claim.
If you want the legal background in more depth, we have written about whether it is legal to record a business meeting with AI and how to get consent to record a business call.
Nothing in the reporting we reviewed says AI notetakers are unlawful as a category. The disputes are about notice, consent and what a vendor does with recordings, and those depend on where the people on the call are located. Treat consent as a requirement and a visible notice as the starting point.
Not on its own. It tells people that recording is happening, which is necessary, but consent can still require that they are asked and able to decline. The safest practice is to say it aloud at the start and stop if someone objects.
Many security teams argue for governed use rather than bans, because a ban sends people to tools nobody has reviewed. Our policy template shows what governed use looks like in practice.
It changes who sees the tool, not who has to agree. Without a bot, nobody on the call sees any notice unless you give one yourself. If you record a browser tab, tell people first, every time.
Yes, for one person, with no card required. You can try it on a real meeting and check the write-up before deciding anything. See pricing.
Try Notepik on your next meeting
Free to start, no card required. Connect a calendar or paste a link, and the summary is waiting when the call ends.
Notetakers spread through calendars faster than IT can review them. A one-page policy settles which tools are allowed, what people are told, who sees the notes and how long they stay. Here is a template.
Evaluating the most secure AI meeting transcription for enterprises requires a deep dive into data protection. Learn what security features matter most.
Evaluating an AI meeting assistant with on premise data options? Understand the critical security, encryption, and data handling considerations for enterprise adoption.