Turn off training and set retention before Otter records company meetings. Otter's privacy policy lets it train on de-identified audio and transcripts unless each user picks Don't allow sharing under Help improve Otter, and only Enterprise workspaces start opted out. Otter keeps audio, transcripts, summaries, and screenshots on AWS in the US West region until someone deletes them, unless Enterprise sets a retention policy.
When you need this
- Security review asks where recordings and transcripts live, for how long, and who can learn from them.
- Customer, candidate, or patient meetings must stay out of vendor model training.
- A healthcare team, or one subject to biometric or EU data rules, is evaluating Otter.
Evaluation criteria
What is stored
Which parts of a meeting does Otter keep?
Look forOtter keeps audio recordings, transcripts, summaries, action items, and virtual meeting screenshots on AWS S3 in the US West region, with AES-256 server-side encryption. Enterprise can remove audio from past and future conversations and keep the transcripts.
How long it is kept
Who sets the limit on retention?
Look forOn Enterprise, ask your account manager to set retention of 24 hours or more, for whole conversations, audio and video, or transcripts. Turning it on deletes every existing conversation older than the limit at once. Those deletions skip Trash, and admins cannot see a list of what was deleted, so export what you need first. The deletion job runs daily and finishes within 48 hours. Deleting transcripts stops summaries from being regenerated. Without a policy, conversations stay until someone deletes them, and a deleted conversation sits in Trash for 30 days unless someone empties it.
Model training
Does our meeting data train anyone's models?
Look forSelect Don't allow sharing under Help improve Otter. It removes Otter's access to past conversations and keeps future ones out of training. Until a user chooses, collaborators can opt shared conversations in. Otter's policy allows training on de-identified audio and transcripts, and its terms also allow aggregated or de-identified data for machine learning. Enterprise workspaces start opted out. AI Chat's outside providers do not train on user data. Otter's subprocessor list, effective March 31, 2026, names Anthropic for AI features and OpenAI for evaluating Otter's language models, and bars both from training on customer data. It also names Research Transcriptions, which annotates Otter's training and evaluation data.
Voice data
Does Otter build voice profiles of participants?
Look forEnterprise customers can have their account manager turn off speaker learning if policy requires it. Otter learns voices from speaker tags and shares the profiles across the workspace. With learning off, Otter ignores existing enrollments, and only retention deletes them.
Compliance
Does Otter meet our regulatory baseline?
Look forRequest the Trust Center reports for SOC 2, HIPAA, GDPR, CCPA, VPAT, and TX-RAMP Level 2. HIPAA needs Enterprise and a signed business associate agreement (BAA). For healthcare use, follow Otter's guidance to control auto-join and turn off public link sharing. Otter stores data in the US. Every subprocessor on its list is in the US except Crescendo, whose support staff in the Philippines and Honduras see customer data only in support cases a customer opens.
People who are not your users
What happens when a participant asks Otter to delete a recording of them?
Look forName an owner for deletion requests about your organization's recordings. Otter sends requests to the conversation owner. A European participant who sent Otter a GDPR deletion request reports that Otter pointed them to that owner.
Questions to ask vendors
- Will you confirm in writing that our workspace is excluded from training on de-identified data, as well as from conversation access?
- Which subprocessors on your list receive audio or transcripts, and does Research Transcriptions ever see our workspace's data?
- Can you list the conversations a new retention policy will delete before we turn it on?
- Can you turn off speaker learning and delete existing voice enrollments for our workspace?
Red flags
- Users on Business or Pro who never set Help improve Otter, recording customer meetings.
- Protected health information recorded outside an Enterprise workspace with a signed BAA.
- An EU data residency requirement, which the documented AWS US West storage does not meet.
- Assuming deleted conversations disappear at once, when they stay in Trash for 30 days.
Products in this piece
Sources
- Official documentation
- 13
- Public source
- 3
- Otter.ai Privacy Policy (effective June 16, 2026)
- Otter.ai Terms of Service (effective September 19, 2025)
- Otter Help Center: Provide feedback to Otter
- Otter Help Center: Enterprise Admin Controls Overview
- Otter Help Center: Set a custom Data Retention policy
- Otter Help Center: Delete a conversation
- Otter Help Center: Data security and privacy policies
- Otter Help Center: Speaker Identification Overview
- Otter Help Center: Disable Speaker Learning
- Otter Help Center: HIPAA | Otter.ai
- Otter Help Center: Otter AI Chat FAQs
- Otter.ai Trust Center
- Hacker News comment on a GDPR deletion request to Otter (Aug 18, 2025)
- Hacker News comment on leaving a confidential meeting Otter joined (Aug 18, 2025)
- VCU Technology Services: Otter.ai is blocked and can't be reached (last updated Jun 12, 2026)
- Otter.ai: Subprocessors (effective March 31, 2026)