Models & training

This doc walks through how we use various AI models and approach training those models at Granola.

How does Granola handle model training?

By default, we may use anonymized data to improve our services, but this data is never sent to third parties.

All users (on any plan) can opt out of model training in their settings. For Enterprise customers, admins can enforce this opt-out across the entire organization with a single setting. We have enterprise agreements with all of our service providers that prevent your data from being used in their model training.

How is data anonymized for model training?

• No third parties train on your data. We have enterprise agreements with all of our service providers that prevent your data from being used in their model training.

• For non-Enterprise users, Granola may use anonymized data to improve our services. However, all users (on any plan) can opt out of model training in their settings.

• For Enterprise customers, Granola opts-out your entire organization from model training.

Which AI models does Granola use?

Summary notes

Granola uses a combination of models for different aspects of the service. We continuously evaluate and switch between models to provide the best experience for different meeting types and use cases.

Ask Granola

When using Ask Granola to chat with your meetings for insights, you can specify the model you'd like to use.

​

Can I use my own AI models with Granola?

Currently, Granola uses our carefully selected models for transcription and summarization. We don’t support bringing your own models at this time, though we’re evaluating this for future enterprise features.