The Granola Lawsuit, Explained

A proposed class action accuses the AI note-taker Granola of recording and using private conversations without everyone's consent. The case is Chamberlain v. Granola, Inc., No. 3:26-cv-07926, filed July 30, 2026 in the Northern District of California.

The suit is early-stage. Every point below is an allegation. No court has ruled that Granola broke any law, and Granola has not yet answered the complaint. Treat this as a live dispute, not a finding.

This article is general information, not legal advice. If your team records meetings, talk to qualified counsel about your own facts before relying on anything here.

Who Sued Granola, and Where?

A named plaintiff, Chamberlain, sued Granola, Inc. on behalf of a proposed nationwide class of people whose conversations the app allegedly captured. The complaint was filed July 30, 2026 in federal court for the Northern District of California.

The plaintiff is described as a Florida resident. That detail matters later, because it shapes which state consent laws can anchor a nationwide class.

The case sits alongside parallel privacy class actions filed against other AI note-takers, including Otter.ai and Fireflies. The legal theory in each is closely related.

As a proposed class action, the suit asks the court to let one plaintiff represent many people at once. That request has to be granted before the case can grow, and it is a separate step from deciding whether Granola did anything wrong.

What Is the 'Silent Capture' Model at the Center of the Case?

Granola is an AI notepad that listens to a meeting through your device and enhances the notes you type. Unlike many rivals, it does not send a visible bot into the call.

That design is the product's selling point and, the complaint argues, its legal problem. When no bot appears in the participant list, other people on the call may never learn that software is capturing what they say.

The complaint alleges this silent, no-bot model means non-consenting participants are recorded without notice. In our AI workflow audits for teams rolling out meeting-capture tools, the failure mode we see most is exactly this: the tool works so quietly that only the person who installed it knows it is on.

What Does the Complaint Say About AI Training?

The complaint alleges that Granola uses captured communications to train AI models by default, rather than only when a user opts in. In the plaintiff's telling, private conversation becomes model input without clear, per-participant permission.

The complaint also raises a practical concern about permanence. It alleges that once data is baked into a trained model, it cannot be cleanly extracted or deleted. Under that theory, a later opt-out would not undo what the model already absorbed.

That permanence argument does real work in the case. If deleting an account cannot pull a conversation back out of a trained model, then the usual privacy remedy of deletion is worth less, and the plaintiff can argue the only real protection was consent up front.

These are contested allegations. Granola has not admitted them, and the technical claims about model extraction will likely be disputed by experts on both sides.

What Is the Federal Wiretap Theory?

The core federal claim is under the Electronic Communications Privacy Act, the modern Wiretap Act. It bars intentionally intercepting the contents of an electronic communication without proper authorization.

The plaintiff's theory casts Granola as an uninvited listener on private calls. If the software intercepts conversation contents in real time, the argument goes, that interception can trigger the statute.

A win under ECPA is powerful because it is federal. It can reach conduct across many states at once, which is part of why plaintiffs anchor these note-taker suits to it.

The statute also reaches more than the vendor. It can touch anyone who procures the interception, which is why the person who installs the tool, and the employer who tells them to, can be drawn into the exposure alongside Granola.

Why Is One-party Consent a Hurdle for the Plaintiff?

ECPA allows interception when one party to the conversation consents. The person who installs Granola is a party to their own meeting, and by turning the tool on, they arguably consent.

That single consent can defeat the federal claim. Plaintiffs try to get around it with the statute's 'criminal or tortious purpose' exception, which withdraws the consent shield when interception serves an unlawful aim.

The Ninth Circuit reads that exception narrowly. Courts there have generally required that the unlawful purpose be the reason for the interception, not merely that the interception itself might amount to a tort. District courts in California have split on how to apply it, which makes the outcome hard to predict.

How Do California's CIPA Claims Fit In?

The complaint also raises California state claims under the California Invasion of Privacy Act, or CIPA. Unlike federal law, California is an all-party (two-party) consent state.

CIPA generally requires that every participant consent before a confidential conversation is recorded. That standard is harder for a silent-capture tool to meet, so state claims often carry more force than the federal count.

The tradeoff is reach. CIPA protects communications tied to California, so it does not automatically cover a nationwide class of users in one-party states.

Why Is Nationwide Class Certification Difficult Here?

Consent law is not uniform. Roughly a dozen states require all-party consent, while the rest and federal law require only one party. A class spanning all of them mixes people under very different rules.

The named plaintiff is described as a Florida resident, and Florida is an all-party state. A plaintiff's own state law shapes the claims they can lead, and courts ask whether one representative can fairly stand in for members governed by other statutes.

When state laws differ this much, defendants argue those differences overwhelm the common questions a class needs. That fight over one-party versus all-party consent is often where these cases are won or lost, well before any trial.

Plaintiffs have options to manage the problem. They can narrow the class to all-party states, split it into subclasses by consent rule, or lead with the federal ECPA count, which applies the same everywhere. Each choice trades breadth for a cleaner class, and the judge's view of that tradeoff will shape how large the case can grow.

What Does This Mean for Users and for Companies Using Granola?

For individual users, the case is a reminder that turning on a silent note-taker may put legal exposure on you, not just the vendor. You are usually the consenting party, which can make you the person who owed notice to everyone else on the call.

For companies, the risk is operational. Recording a customer or candidate in an all-party state without clear consent can create wiretap exposure regardless of how the vendor's lawsuit ends.

The safest path does not wait for a ruling. It builds consent into the workflow: announce capture, get agreement on the record, and turn training defaults off where you can. We help teams put those controls in place before a tool is switched on company-wide.

• Get explicit consent from every participant, especially in all-party consent states.
• Check whether your note-taker trains on your data by default, and change the setting if you can.
• Write a recording-notice step into your meeting process instead of relying on the tool to disclose itself.
• Keep a record of consent for external calls with customers, candidates, or vendors.