Does Granola train AI models on your meetings? Model training policy by plan

TL;DR: We never let third-party AI providers train on your meeting data. That is contractually prohibited on every plan. We do use anonymized data to improve our own models internally, and this is on by default on Basic (Free) and Business plans. You can opt out at any time under Settings > Preferences > Data & sharing. The toggle is labeled "Use my data to improve models for everyone." Enterprise plans default to model training off org-wide, with admins managing the setting at the workspace level.

Questions about AI model training come up early when teams evaluate any tool that touches customer conversations. This article breaks down our model training policy by plan, where to find the settings, and what the architecture means in practice.

Enterprise defaults to model training off for the whole org

We set the Enterprise plan to opt out of model training org-wide by default. This means no user data from Enterprise accounts trains our AI models unless an admin explicitly enables it.

Internal training means improving our own models, while third-party training means external providers using your transcripts to train their public models.

We earned SOC 2 Type 2 certification in July 2025 and you can view the full report at trust.granola.ai.

Enterprise customers also get configurable data retention windows (configured by contacting Granola), SSO/SAML integration, and audit APIs. These controls let IT teams enforce compliance requirements without relying on individual user settings.

Individual users don't need to remember to change anything. Research teams using Enterprise accounts are covered by the org-wide default.

Basic and Business plans enable internal training by default

Basic and Business plans default to internal model training enabled. This means we use anonymized data from your meetings to improve our AI models and product features.

You can opt out of internal training on any plan, including Basic.

Plan comparison table

Third-party training is prohibited on every plan. The only variable is whether we use anonymized data to improve our models, and whether that default is on or off.

What the data privacy toggle does

The opt-out setting controls internal model training. When enabled (default on Basic/Business), we use anonymized meeting data to train our AI models. When you disable it, we don't use your data for any training.

When internal training is enabled, we use anonymized meeting data only.

What the architecture does with your audio

Our privacy architecture makes several commitments about how your audio is handled:

• Audio deletion after transcription: We temporarily cache audio during the meeting for transcription only, then delete it once transcription completes.
• Anonymization: When internal training is enabled, we strip identifying information before use.

How Granola handles your meeting data

Here is how it works:

1. Audio capture: We capture device audio during the meeting
2. Real-time transcription: We transcribe audio as the conversation happens
3. Audio deletion: We delete the cached audio once transcription completes
4. Note storage and enhancement: We store transcripts and your notes in the cloud, and our AI processes your notes to add context from the transcript

We prioritize privacy over features like audio playback. When transcripts are auto-deleted, only the underlying transcript is removed. Your enhanced notes and any rough notes you've taken will remain available. You can also delete your account at any time from Settings > Profile > Delete my account. Notes you own are permanently removed.

You can opt out of internal training on every plan

You can check your current training status in your settings.

On Basic or Business plans:

1. Open Granola and go to Settings
2. Navigate to Preferences
3. Select Data & sharing
4. Look for the toggle labeled "Use my data to improve models for everyone." If it's on, internal training is enabled. If it's off, you've opted out.

On Enterprise plans, the toggle is labeled 'Use my workspace data to improve models for everyone' and is managed at the workspace level. Admins can change it themselves in Settings > Workspace > General > Data security. The default is off.

If you're evaluating Granola for your team, you can test this on the Basic plan. Create a free account, run a meeting on the Basic plan, and check the settings location to confirm you can find and control the toggle.

Encryption, access controls, and audit logging

Beyond training policies, our architecture limits data exposure:

• Encryption: We encrypt data in transit and at rest
• Access controls: Only you can access your notes unless you share them
• Audit logging: Enterprise plans include audit APIs to track data access
• Data residency: We host your data in US-based AWS infrastructure, consistent with our broader security practices
• GDPR compliance: We comply with GDPR and offer a Data Processing Agreement

These protections are in place whether or not internal training is enabled. Anonymized data used for training is never sent to third parties, and our service providers cannot use your data for their own model training.

Admin controls versus individual opt-out: What each plan gives you

The key difference between plans is who controls the settings, not how we handle your data.

Individual opt-out settings give you direct control over internal training, and notes can be deleted manually at any time.

Enterprise plans give admins org-wide control. We turn training off by default, let admins enforce retention windows, and let admins manage user access through SSO and SCIM.

For teams handling sensitive participant data, Enterprise controls centralize this at the admin level rather than relying on individual user settings.

FAQs

Can I turn off model training on Basic and Business plans?

Yes. Go to Settings > Preferences > Data & sharing, toggle off "Use my data to improve models for everyone," and we stop using your meeting data for internal training.

What happens to my meeting data if model training is enabled?

We use anonymized data from your meetings to improve our AI models and product features after stripping identifying information. Your data does not train third-party public models.

Does Granola train on my data even if I disable the setting?

No. When you opt out of model training, we stop using your meeting data for internal training. We prohibit third-party providers from training on your data regardless of this setting.

How does this compare to other AI notepads?

Some AI notepads retain audio for playback. Granola does not. You can't replay the exact tone of voice from a conversation. What you do have is the full transcript and source-linked citations in Granola Chat, so you can verify exact quotes and surface patterns across past interviews without needing the audio file. On third-party training, policies vary across tools. Check each vendor's privacy documentation directly before committing to a tool.

Where can I find Granola's full security documentation?

Our security page covers SOC 2 certification, GDPR compliance, data retention, and third-party AI policies.

Key terms glossary

Model training: The process of using data to teach an AI system to improve its predictions or outputs. Internal model training refers to a company using data to improve its own AI models, while third-party training refers to external AI providers using data to train their public models.

Third-party AI providers: External companies that supply large language models and other AI capabilities used by software products for features like text generation, analysis, and enhancement.

Data retention: The policies and practices governing how long a company stores user data before deleting or anonymizing it. Retention periods vary by data type, regulatory requirements, and business needs.

Anonymized data: Data from which identifying information has been removed to reduce the risk of re-identification.

SOC 2 Type 2: An independent audit certification verifying a company's controls over security, availability, processing integrity, confidentiality, and privacy over a defined period. The SOC framework is developed by the American Institute of CPAs (AICPA), and reports are issued by independent CPA firms. SOC 2 Type 2 is widely recognized as an industry standard for service organizations.