A proposed class action filed July 30, 2026 in the U.S. District Court for the Northern District of California, docketed as Case No. 3:26-cv-07926-EMC, accuses AI meeting-notetaking company Granola of recording conversations without most participants’ knowledge and using the recordings to train its own AI models. The suit names Delaware corporation Granola, Inc. and UK-registered Granola Labs Ltd. as defendants. Plaintiff Tarra Chamberlain, a Florida resident, filed on behalf of a proposed nationwide class and a California subclass; attorney Rob Freund made the complaint public on August 1.
What Is the Granola Lawsuit About?
The Granola lawsuit alleges that the company’s AI notetaker captures microphone and system audio during virtual and in-person meetings without disclosing this to most participants, then uses those recordings by default to train Granola’s own AI models on the Free and Business plans. The complaint asserts seven causes of action under federal and California privacy law, and seeks statutory damages of $5,000 per violation. These are allegations in a newly filed complaint: no court has ruled on their merits, and as of the PPC Land report neither defendant had publicly responded.
What the Complaint Alleges
According to the filing, Granola’s notetaker captures audio through a participant’s microphone and the device’s system output, then routes the recording to a transcription vendor. The complaint claims the resulting data is used, by default, to train Granola’s own models, a setting the plaintiff says goes undisclosed to most participants. According to the filing, switching it off requires a manual change by the user.
The complaint also quotes Granola’s own marketing copy in support of that allegation:
“Other people in the room won’t know it’s there.”
— Granola marketing copy, quoted in the complaint
The plaintiff argues that framing undercuts the disclosure needed for consent.
How Granola Describes Its Own Consent Practices
Granola’s help center documentation, published separately from the litigation, strikes a different tone: “It is best practice to make sure participants know when you use AI notetaking tools.” For virtual calls, the same documentation describes on-screen signals: “When enabled, Granola shows a watermark on the user’s camera and an automated message in the meeting chat that are visible to everyone.”
The documentation separately states: “Our agreements with sub-processors prohibit them from using your data for model training.” That commitment covers third-party vendors handling data on Granola’s behalf. It does not address whether Granola itself trains its own models on customer meeting content, which is the practice the complaint challenges.
Seven Legal Claims, One Underlying Question
The complaint lists seven causes of action, each built on the allegation that Granola and Granola Labs recorded conversations without adequate consent:
- Invasion of privacy (intrusion upon seclusion)
- Violation of the federal Electronic Communications Privacy Act
- California Penal Code § 631
- California Penal Code § 632
- California Computer Data Access and Fraud Act
- California Unfair Competition Law
- Unjust enrichment
The complaint seeks $5,000 per violation or three times actual damages under Penal Code § 637.2(a). It describes a proposed class reaching millions of people and an alleged amount in controversy exceeding $5,000,000; no damages have been awarded.
What This Means for Teams Running Meeting Notetakers
The suit targets a vendor, not the employees running the tool on client calls. Granola’s own recommendation, that participants should know an AI notetaker is active, is worth checking regardless of how the litigation ends:
- Are the notetaker’s indicators (watermark, chat message, spoken disclosure) visible to every participant, not just switched on in an admin panel?
- What does the vendor’s agreement say about training its own models on meeting content, as distinct from what it promises about sub-processors?
- Does the default configuration match what the vendor’s own documentation tells customers to do?
The same question extends past audio. A separate indexing failure that put shared AI chat transcripts into Google’s search results showed how one overlooked default can expose content participants assumed stayed private. For companies mapping obligations under California’s mandatory risk assessments and rules on automated decision-making tools, a notetaker facing a wiretapping suit over its alleged recording defaults belongs on the vendor review list.
What Happens Next
As of the PPC Land report, neither Granola, Inc. nor Granola Labs Ltd. had publicly responded. The allegations above come from the plaintiff’s complaint, not a court finding, and all seven claims remain to be tested.
Sources: PPC Land; Granola’s own consent documentation.