AI notetaker lawsuits in 2026: the bot was never the legal problem

The AI notetaker lawsuit everyone saw coming arrived in August 2025, when four class actions landed on Otter.ai over a bot that joined meetings and recorded them. The industry read that as a verdict on bots, and the fix looked obvious: build a notetaker that doesn't send anything into the call. Then in July 2026 the most prominent bot-free notetaker was sued too — and the complaint's central exhibit was its own marketing promise that other participants "won't know it's there." Three AI notetaker lawsuits are now live against three of the category's best-known products, and read together they say something the industry's risk narrative got backwards. The exposure was never really the bot. It was consent and what happens to your meeting afterwards.

This piece is a factual map of who is being sued, for what, and what has actually been decided so far. Every case below is unresolved. Allegations in a complaint are allegations, and a judge letting a claim proceed is a ruling about whether a case may go forward, not a finding that anyone did anything wrong. We are a tools directory, not a law firm, and none of this is legal advice.

Where the AI notetaker lawsuits stand

Otter: the case that survived the motion to dismiss

Otter's is the furthest along, and the only one where a judge has weighed in on the substance.

In re Otter.AI Privacy Litigation consolidates four putative class actions — Brewer, Walker, Theus and Winston — filed between August and September 2025 and combined before Judge Eumi K. Lee in the Northern District of California on October 22, 2025. The consolidated complaint landed December 5, 2025. The core allegation is that OtterPilot joined Zoom, Teams and Google Meet calls, then recorded, transcribed and retained what was said without the consent of everyone present, and that the recordings were used to improve Otter's systems. The plaintiffs who matter most to the theory are the ones who never signed up for anything: people who were simply in a meeting somebody else had configured.

On August 13, 2026, Judge Lee granted Otter's motion to dismiss only in part. What survived is the serious half: the federal Wiretap Act claim, the California Invasion of Privacy Act claim, and both Illinois biometric-privacy claims, on the theory that Otter plausibly acted as a third-party eavesdropper rather than a mere tool of the meeting host. What went was the computer-intrusion framing — both CFAA counts, the California CDAFA claim, the Washington Privacy Act claim and most of the intrusion-upon-seclusion claims, dismissed with leave to amend.

That distinction is the whole story of the ruling. The court was unpersuaded that a notetaker is like hacking a computer. It was persuaded, at the pleading stage, that a notetaker might be like a stranger listening to a call. Discovery now runs on that question.

Granola: the bot-free tool sued for being invisible

Granola built its reputation on the opposite design. Nothing joins the meeting; the app runs locally, listens to your computer's audio, and expands the sparse notes you type yourself. For two years that was pitched — by the company and by people like us writing about it — as the answer to bot fatigue and the consent problem alike.

Chamberlain v. Granola was filed July 30, 2026 in the Northern District of California, the same court hearing the Otter case. The complaint alleges Granola recorded a meeting participant without notice that an AI notetaker was present at all, and that meeting content was used by default for commercial purposes including AI model training unless a user went and turned that setting off. It pleads violations of the federal Electronic Communications Privacy Act, the California Invasion of Privacy Act, California's confidential-communications statute, common-law invasion of privacy, and California's computer-access statute.

The detail worth sitting with is what plaintiffs' counsel reached for as evidence of intent: Granola's own website copy, promising that other participants "won't know it's there." A sentence written as a feature became the argument that the concealment was designed rather than incidental.

There is a real distinction the complaint draws between Granola and its bot-based competitors, and it does not run the way the category assumed. A bot that joins a call is at least visible. Participants can see it in the attendee list, ask about it, and in many products remove it. A tool that captures the same audio while announcing nothing removes the one signal that prompted anybody to object. On the specific question of notice, invisibility is a liability, not a mitigation.

None of this has been decided. Granola may well prevail. But the argument that going bot-free settles the consent question did not survive contact with a docket.

Fireflies: the biometric claim nobody priced in

Fireflies.ai faces a different theory, and one that is easy to miss because it has nothing to do with whether a recording happened.

Cruz v. Fireflies.AI Corp. (3:25-cv-03399) was filed in Illinois in December 2025 under the Illinois Biometric Information Privacy Act, followed by a second case in the Northern District of Illinois. BIPA is the statute that has produced the largest privacy settlements in the United States, and it does not care about wiretapping. It cares about biometric identifiers.

The claim is that Fireflies' Speaker Recognition feature — the thing that labels who said what — necessarily builds voiceprints, and that voiceprints are biometric identifiers expressly covered by BIPA. On that reading, the statute's requirements attach: a published retention and destruction schedule, written notice of what is collected and why and for how long, and a written release from each person whose voiceprint is captured. Including, again, people who never held an account and were simply present on a call.

Speaker labelling is a table-stakes feature. It ships in most of this category. If the theory holds, the exposure is not specific to Fireflies at all.

Who isn't being sued — and why that isn't a safety rating

As of August 28, 2026 we could find no privacy case against Fathom or Read AI. That is worth stating plainly, because a list of defendants implies a list of non-defendants, and people will read the second list as a recommendation.

It should not be read that way. Not being sued is not a compliance certification. It reflects who plaintiffs' firms have targeted so far, which tracks brand recognition, user volume, jurisdiction and the specific features a statute happens to bite on — not a considered judgement about which products handle consent best. Both cases in Illinois exist because Illinois has BIPA, not because Illinois notetakers behave worse. A tool with a smaller footprint in a state with a private right of action is a tool that has not been tested yet.

What the three cases have in common

Strip the three complaints down and the same two things appear in all of them, and neither is the bot.

Nobody asked the room. Every one of these cases turns on participants who did not consent, usually because nobody put the question to them. That is a design and workflow problem, and it belongs as much to the organisation running the tool as to the vendor selling it. Many US states require all-party consent to record a conversation; the meeting host is the person in a position to obtain it, and no product does that for you.

The default did the deciding. Two of the three complaints allege that meeting content fed model training unless the user found and changed a setting. The recording is the visible part of the transaction. What happens to it afterwards is the part with the statutory damages attached, and it is usually configured once, by someone who never looked.

The category spent two years treating "does it send a bot?" as the risk question because it was the visible difference between products. It was the wrong variable. The right ones are whether everyone in the meeting knew, and whether the transcript trains somebody's model.

What this means if you already run one

The practical steps are unglamorous and mostly not about which product you bought.

Tell people before recording starts, out loud or in the invite, and give them a way to say no. Find the model-training setting in whatever you use and decide deliberately rather than inheriting a default — Granola, for instance, lists a model-training opt-out on its free tier and a team-wide opt-out at the Enterprise level, and the equivalent exists in most competitors. Work out whether anyone you meet with sits in an all-party-consent state or in Illinois, because that changes your exposure regardless of vendor. Decide which meetings simply should not be captured; the fastest-growing practice in the category is teams keeping a channel where nothing is recorded at all. And if you are in a regulated function or handling HR conversations, ask counsel rather than a directory.

What this means if you're choosing one

Choose on fit. The litigation does not sort these products into safe and unsafe, because the exposure follows how you deploy a tool far more than which logo is on it.

What has genuinely changed is a marketing claim you should now discount. "No bot in the call" is a real feature with real benefits — less meeting clutter, no third participant, in-person capture, a better experience for the person actually in the room. It is not a consent shield, and the first court test of that proposition came from suing the bot-free product. Buy bot-free because you want the workflow. Do not buy it because you think it answers the legal question.

Our full head-to-head on features, pricing and fit is the companion read: AI notetakers in 2026: Otter vs Fireflies vs Granola vs Fathom vs Read. For the broader ranked field, see our Top 7 AI Meeting Note-Takers collection.

Frequently asked questions

Which AI notetakers are being sued in 2026? Three of the best-known products face privacy litigation. Otter.ai is the defendant in In re Otter.AI Privacy Litigation (5:25-cv-06911-EKL, N.D. Cal.), where a judge allowed federal wiretap, California privacy and Illinois biometric claims to proceed on August 13, 2026. Granola was sued in Chamberlain v. Granola, filed July 30, 2026 in the same district. Fireflies.ai faces Cruz v. Fireflies.AI Corp. (3:25-cv-03399) and a second Illinois case, both under the Illinois Biometric Information Privacy Act. We found no case against Fathom or Read AI as of August 28, 2026. All three cases are unresolved.

What is the Otter.ai lawsuit about? It alleges that Otter's OtterPilot bot joined Zoom, Microsoft Teams and Google Meet calls and recorded, transcribed and retained conversations without the consent of every participant, and that recordings were used to improve Otter's systems. Four class actions filed in August and September 2025 were consolidated before Judge Eumi K. Lee in October 2025. On August 13, 2026 the court dismissed the computer-intrusion claims with leave to amend but allowed the Wiretap Act, California Invasion of Privacy Act and Illinois biometric claims to move into discovery.

Why was Granola sued if it doesn't use a meeting bot? Because the case is about notice, not bots. Chamberlain v. Granola alleges Granola captured a participant without any indication an AI notetaker was present, and used meeting content for AI training by default unless a user disabled it. Plaintiffs' attorneys pointed to Granola's own marketing that other participants "won't know it's there." A bot at least appears in the attendee list; a tool that announces nothing removes the signal that would prompt someone to object, which is why bot-free capture does not resolve the consent question.

Is it illegal to use an AI notetaker? Using one is not illegal. The contested questions are whether recording without every participant's consent violates federal or state wiretap law, and whether generating voiceprints without written notice and release violates biometric statutes like Illinois's BIPA. Many US states require all-party consent to record. The obligation to obtain that consent generally sits with the person running the meeting, not the software vendor, and no court has yet issued a final ruling on the merits in any of these cases.

Do AI notetakers train on my meetings? Often by default, though most now offer a way to turn it off. Two of the three 2026 complaints allege exactly this — that meeting content fed model training unless the user found and changed a setting. Check the setting in whatever tool you use rather than assuming; it is usually configured once, at signup, by someone who did not read it.

Is a bot-free AI notetaker safer legally? No, and 2026 is the year that stopped being arguable. The first lawsuit against a bot-free notetaker turned on the very invisibility that was sold as the benefit. Bot-free capture has genuine advantages — less clutter, in-person meetings, a better experience for attendees — but consent is a workflow obligation the product cannot discharge for you.

What should my team do about AI notetaker privacy? Announce recording before it starts and allow people to decline; review the model-training setting in your tool and set it deliberately; identify whether participants are in all-party-consent states or in Illinois, where BIPA carries a private right of action; and designate conversations that are never captured. If you handle HR, health, legal or financial matters, take advice from counsel rather than from a comparison page.

Where to go next

For features, pricing and which tool actually fits your team, read the head-to-head: AI notetakers in 2026: Otter vs Fireflies vs Granola vs Fathom vs Read. The individual profiles carry current pricing and plan detail: Otter, Granola, Fireflies, Fathom and Read AI. For the wider ranked field, see Top 7 AI Meeting Note-Takers. And if you are weighing how durable any AI tool is before building a workflow on it, our guide to telling whether an AI tool is legit is the companion piece.

We will update this page as these cases move. The next scheduled milestones are discovery in the Otter case and responsive pleadings in Granola.

— The ToolDirectory.AI editorial team