Otter's general privacy policy permits training its proprietary AI using de-identified recordings and transcripts. Enterprise workspaces have a different documented default: they are opted out of model training. The documented training/feedback control also lets an account revoke access. Identify the workspace, agreement, and permission setting covering your meeting.
The relevant sources are Otter's privacy policy, effective June 16, 2026, and its Enterprise admin guide. We checked them on September 22, 2026. We make Anarlog; this article explains published policy rather than auditing Otter's training systems.
Four statements that refer to different things
| Statement | What it covers | What it does not establish |
|---|---|---|
| Otter may improve its proprietary AI with de-identified content | Otter's general training practices | The terms of every Enterprise workspace |
| Enterprise is opted out by default | The documented Enterprise agreement and configuration | An identical default on Basic, Pro, or Business |
| Third-party AI providers cannot train on customer data | Otter's commitments concerning those providers | Whether Otter itself trains on the content |
| Human review requires consent | Manual access described in the security FAQ | An opt-out from automated training |
The third and fourth distinctions come from Otter's Privacy & Security FAQ. Reading only the answer about external model providers can miss the separate answer about Otter's own models.
De-identification is not the same as opting out
Otter describes removing identifying information before automated training. Its privacy policy nevertheless identifies recordings and transcripts among the inputs used to improve its technology. That is a statement about permitted processing, not evidence that a particular meeting was included in a training run. Source: Otter privacy policy.
For a team deciding whether to use the product, the useful requirement is precise: "Our meeting content must not be used to improve shared models." That differs from "remove names before training" and from "do not let a person read the transcript."
Write down the requirement before comparing tools. Otherwise, a security answer can sound reassuring while addressing a different question.
What Enterprise changes
Otter's Enterprise documentation says the no-training default covers recordings, transcripts, and notes used to provide the workspace's features. The setting is managed by Otter under the Enterprise arrangement; the guide directs customers to their account manager to opt in. Source: Enterprise controls.
Do not infer the same protection from a paid subscription alone. Confirm the exact workspace containing the meeting and the applicable agreement. A person can use a work email address without establishing that every recording belongs to the intended Enterprise workspace.
Change training and feedback permissions
Otter documents Account Settings → Meetings → Feedback and Training → Help improve Otter. Select the restrictive sharing option to revoke conversation access for training and product improvement. Its feedback guide explicitly says this removes access to past conversations and excludes future conversations from training.
Source: Otter.
Read feedback prompts too: granting the optional permission there can change the account setting back to allow access. Simply undoing a feedback rating does not reverse that account-level choice.
Source: Otter.
There is also a collaboration exception. When the owner's account preference is unset, a collaborator can grant training access for an individually shared conversation through feedback. Setting the owner's preference explicitly prevents collaborators from making that opt-in. These are documented controls; we did not toggle a live Otter account for this article.
The guide does not establish that revoking access reverses a training run already completed. Keep that separate from its promise about access to past conversations and future use.
Source: Otter.
Imported documents and third-party AI are separate cases
Otter's security FAQ makes a separate commitment concerning imported customer data, including documents from Google Workspace. It also describes restrictions on external AI service providers. Do not automatically extend a statement about imported Workspace documents to every audio recording created in Otter. Source: Otter security FAQ.
The same distinction matters when a meeting tool offers several features. Recording audio, connecting a document source, and asking an AI chat question may involve different inputs and processors. Check the feature you intend to use.
A short account check before recording
Use this checklist to obtain an answer you can rely on:
- Identify the plan, workspace, and account that will own the recording.
- Confirm whether the general policy or an Enterprise agreement governs that content.
- Ask whether audio, transcripts, notes, feedback, and connected documents are covered by the same training restriction.
- Check retention and sharing independently. No training does not mean no storage or no teammate access.
- Keep the dated policy and account-specific confirmation with your team's tool decision.
This is a product-configuration checklist, not a legal approval. Recording consent and your organization's information policy still apply.
If you want a local AI workflow
Otter can be a practical choice for a team that wants a hosted meeting workspace, especially when its Enterprise controls match the team's requirements. The relevant comparison is the actual configuration, not a claim that every cloud product behaves identically.
Anarlog keeps canonical meeting data in local SQLite and lets you select local or external AI providers. A fully local setup requires both local transcription and local summaries; bringing your own cloud API key still sends the relevant input to that provider. See the offline documentation before using local processing as your reason to switch.
For the wider product assessment, read our Otter review. For the training question, keep the answer scoped to the plan, data type, processor, and agreement you verified.