Is Granola AI safe for client conversations?

How does Granola capture a conversation?

Granola runs on a computer or phone and uses device audio to prepare a transcript. On desktop, that includes microphone and system audio. The app works with common meeting platforms without adding an extra attendee. Its security documentation describes cloud services used for transcription and AI processing.

Granola says it deletes audio after transcription. Desktop transcription runs in real time, while mobile can temporarily cache audio for processing. The resulting transcript and notes have their own retention settings. Opening a future meeting note can start transcription at the scheduled meeting time. Read the security overview and transcription guide.

How can other participants know Granola is running?

Granola offers an automated chat message and a video watermark to tell participants that transcription is running. Individuals can enable these features, and Enterprise administrators can require them across a workspace. The setup depends on the meeting platform, browser, operating system, and permissions granted to the app.

Current guides cover Google Meet, Zoom, and Microsoft Teams on macOS and Windows, with Slack huddles supported on macOS. The watermark requires the Granola camera to be selected. Enterprise Heads Up can add notice before a meeting, after an administrator enables it and configures the calendar connection. Review transparency setup and Heads Up.

A team should test what an external attendee actually sees, including someone who arrives late. Record how the organizer handles an objection and who can stop capture. Your client agreement and applicable consent requirements should guide that procedure.

What happens to meeting data used for AI training?

Granola distinguishes its own model improvement from training by external AI providers. Its documentation says provider agreements prohibit those providers from training on user data. Granola may use anonymized data for its own models by default on Basic and Business, with an account setting that lets users opt out.

Enterprise workspaces are opted out by default. The personal setting applies across the account, including Android when changed through desktop or iOS. The privacy policy says de-identified data previously incorporated into models may remain there. Record the setting before a trial and check who can change it. Read model-training controls and the privacy policy.

How do retention and sharing settings work?

Meeting notes and transcripts can contain client information long after a call ends. A useful review follows that information from capture through sharing and deletion. Granola gives users transcript-retention controls, while Enterprise administrators can apply a workspace policy. The chosen period should match the purpose of retaining the material.

Individual transcript auto-deletion options range from one day to one year, with an Off setting. Changes start a one-week cooldown, and the policy applies to transcripts while keeping notes available. Review shared copies separately. Granola says notes are private until shared through links or folders. Read transcript retention and sharing controls.

Which security details should a firm verify?

Granola reports SOC 2 Type II certification, encrypted data in transit and at rest, and storage on AWS in the United States. These details help a security team frame its assessment. A procurement review should also examine the available report, contractual terms, access permissions, and the specific data the firm expects to process.

Granola offers a Data Processing Addendum. Its documentation describes HIPAA support for Enterprise customers with a Business Associate Agreement, with scope to confirm for the intended apps and use. Ask the vendor for the current subprocessor list and evidence relevant to your requirements. Read the security FAQ and Data Processing Addendum.

What does the Granola lawsuit allege?

On July 30, 2026, Tarra Chamberlain filed a proposed class-action complaint against Granola, Inc. and Granola Labs Ltd. in the Northern District of California. In that original complaint, the plaintiff alleges that Granola captured meeting participants’ communications without their knowledge or consent and used meeting information for model training. Read the original complaint.

Docket Nexus’s public index lists an amended complaint dated October 19, 2026. The account above describes the July 30 filing. See the dated index entry.

Granola's current documentation recommends informing participants, provides transparency controls, and says external model providers are prohibited from training on user data. Read Granola's transparency page and model-training controls.

Which questions belong in your approval checklist?

A client-services firm needs a procedure that people can follow during a real call. Assign an owner to approve the tool and record the settings used in the trial. Then test the experience with a colleague acting as an external participant, including the process for declining capture.

1. Which client agreements permit capture, and which conversations require separate approval?
2. How will participants receive notice and communicate their permission or objection?
3. Who can access the notes, and what happens when someone shares a link?
4. Which training settings apply to this account and workspace?
5. How long will transcripts, notes, and exported copies be kept?
6. Who handles deletion requests, access reviews, and changes to the policy?

How Timeglass identifies speakers

Speaker diarization separates a recording into turns by different people. Timeglass names speakers automatically, separates your voice from the call audio, and recognizes returning voices when Speaker Names is enabled. The names help you check who gave an instruction or made a commitment before sharing a meeting summary.

Timeglass merges teammates’ recordings of the same call into one transcript. Action items stay connected to people, and connected AI tools can use the speaker context. Learn more about speaker identification.

How does Timeglass fit this evaluation?

Timeglass provides meeting summaries, action items, speaker-labelled transcripts, and playback of available video and audio. The speaker timeline shows when each person spoke and their share of attributed speaking time. These records sit alongside the desktop activity used to prepare timesheet entries.

Include video, audio, transcripts, summaries, and downloaded copies in your access and retention review. Timeglass supports view-only meeting links with expiry and revocation controls, while meeting visibility follows workspace permissions and sharing settings. A firm should define who can access each record and how employees control timesheet submission.

Timeglass has an independent SOC 2 Type II audit of its security controls. You can review the published security information with the person responsible for your firm’s requirements. Read Timeglass’s security information.

Timeglass’s privacy policy says it uses paid AI services under terms that prohibit providers from using customer content to train general-purpose models. Provider retention varies, including retention for security, abuse monitoring, or legal obligations. Read Timeglass’s privacy policy.

Test the recording notices participants receive, who can access each meeting record, and the applicable retention settings. Review your client agreements and vendor contract terms before approving the workflow. Include notes and recordings copied into other systems in that review. Review Timeglass capture and timesheet review.

What do teams commonly ask about Granola privacy?

The questions below summarize the distinctions that usually need a clear answer during evaluation. Treat them as a starting point for checking the account and workspace you intend to use. Save the relevant settings and policy dates with your approval record so future reviewers can understand the decision.