Do AI Meeting Note-Takers Train on Your Recordings? Zoom, Fireflies, tl;dv, Granola and Otter
Three of the five AI note-takers PolicyPeek has checked say in their own policies that they don’t train on your meetings: Zoom AI Companion, Fireflies.ai and tl;dv. Granola and Otter.ai train by default but let you opt out. None of the note-takers checked has the ‘no opt-out’ badge.
Last checked: 28 October 2026 (latest check across the tools below; each row shows its own date). Verdicts come from each company’s own published policy, read by PolicyPeek. Not legal advice.
AI note-takers compared
| Tool | PolicyPeek badge | What we recorded from the policy | Source | Last checked |
|---|---|---|---|---|
| Zoom AI Companion | Doesn’t train on your data | “Zoom does not use any of your audio, video, chat, screen sharing, attachments or other communications-like Customer Content (such as poll results, whiteboard and reactions) to train Zoom or third-party artificial intelligence models.” | policy | 2026-10-28 |
| Fireflies.ai | Doesn’t train on your data | We do not use personal information for AI model training and we contractually prohibit our vendors from using this information for their own model training. | policy | 2026-10-16 |
| tl;dv | Doesn’t train on your data | “Tldx Solutions GmbH does not use Customer Content, including meeting recordings, transcripts, notes, files, or other data processed through the Services, to train, fine-tune, or improve foundation models, large language models, or other generative AI models”. | policy | 2026-10-24 |
| Granola | Trains by default, you can opt out | We only use de-identified data to train AI models, which you can opt-out of within your Granola account settings. | policy | 2026-10-15 |
| Otter.ai | Trains by default, you can opt out | Privacy policy: “training our proprietary AI technology on de-identified audio recordings and on transcriptions (which may contain Personal Information)”. Otter’s help centre gives an opt-out: Account Settings > Meetings > Feedback and Training, set Help improve Otter to “Don’t allow sharing”, and “Otter will not access your conversation for training and product improvement.” The policy doesn’t say whether this also covers de-identified training. | policy, policy | 2026-10-24 |
How do I opt out of Granola training?
Granola’s privacy policy says it only uses de-identified data to train AI models, “which you can opt-out of within your Granola account settings”. (Granola)
Can I opt out of Otter.ai training?
Yes, according to Otter’s help centre. Its privacy policy lists “training our proprietary AI technology on de-identified audio recordings and on transcriptions (which may contain Personal Information)”. Its help centre says: “Manage your feedback and training permissions in Account Settings > Meetings. You have the option to allow Otter and its third-party service provider(s) permission to access your conversations for training and product improvement, including through human review.” An admin help page says: “By default, Enterprise workspaces are opted out of AI model training.” None of these pages says what the setting defaults to for individual plans, which plans it covers, or whether it also covers the de-identified training, so the tools above that say they don’t train are the safer choice. (Otter.ai)
The small print
Zoom’s terms say: “Zoom does not use any of your audio, video, chat, screen sharing, attachments or other communications-like Customer Content (such as poll results, whiteboard and reactions) to train Zoom or third-party artificial intelligence models.” The promise is about that Customer Content. (Zoom AI Companion) Otter’s privacy policy mentions training on de-identified recordings and transcripts, and PolicyPeek notes the policy doesn’t say whether the help-centre opt-out also covers that. A recording includes everyone in the call, so this matters for clients and colleagues too.
More: all guides · AI tools that don’t train on your data
How PolicyPeek decides
An AI model extracts the facts from the policy, then fixed rules decide the badge. Results the rules aren’t sure about are held for a person to look at; not every verdict is reviewed by a person. Each tool page links the exact policy used and the date it was checked. (Methodology)
Changes
- 4 November 2026: Otter.ai’s verdict kept on the site (Trains by default, you can opt out) while PolicyPeek re-checks its opt-out wording.
- 28 October 2026: Zoom quote replaced with the word-for-word terms text.
FAQ
Which AI note-takers don’t train on your data?
Of the five note-takers PolicyPeek has checked, Zoom AI Companion, Fireflies.ai and tl;dv have the badge “Doesn’t train on your data”.
Does Otter.ai train on my recordings?
Otter’s privacy policy lists “training our proprietary AI technology on de-identified audio recordings and on transcriptions (which may contain Personal Information)”. Its help centre describes an opt-out under Account Settings > Meetings > Feedback and Training.
Does Granola train on my meetings?
Granola’s policy says: “We only use de-identified data to train AI models, which you can opt-out of within your Granola account settings.”
Does Zoom use my meetings to train AI?
Zoom’s terms say: “Zoom does not use any of your audio, video, chat, screen sharing, attachments or other communications-like Customer Content (such as poll results, whiteboard and reactions) to train Zoom or third-party artificial intelligence models.”
Does tl;dv train on meeting recordings?
No. tl;dv’s privacy policy says it doesn’t use customer content, including meeting recordings and transcripts, to train foundation or large language models.
Does Fireflies.ai train on my meetings?
Fireflies.ai’s policy says: “We do not use personal information for AI model training and we contractually prohibit our vendors from using this information for their own model training.”
Last updated . Plain-text version for AI tools: ai-meeting-note-takers-that-dont-train-on-your-data.md.