Our goal with Griptape is to empower our customers to test and use best-in-class models that align with their own risk tolerance. To provide this flexibility, Griptape supports a wide range of open-source, non-commercial and commercial models. Because AI is a fast-moving space with a wide variety of training data methods, users must verify the safety, compliance, and licensing of each model for their specific use cases. Please note that our supported model availability is subject to change as we continuously evaluate and update our offerings.

 

What data is shared with third-party model providers? Can our data be used to train their models?

How your data is handled by third parties depends on the model and the authentication method you choose within Griptape:

Local Open-Source Models: No data leaves your network and no information is shared with third parties.

Cloud Models via Griptape’s Model Proxy: Griptape routes your data directly to the provider’s API. How that data is utilized by the model provider is governed by the standard terms of service with that specific provider.

Cloud Models via Bring Your Own (BYO) API Key: Griptape acts as a passthrough, routing your data directly to the provider’s API using your authentication details. Use of your data is governed by the commercial agreement your studio holds with that specific provider. 

We strongly advise your security and legal teams to review the policies of all third-party models utilized in your pipeline.

For your convenience, we have provided references for each third-party model’s terms below. Third-party terms are subject to change without notice. While we strive to keep this directory updated from time to time, we strongly encourage all customers to undertake their own due diligence and verify the latest terms and conditions directly with the providers before deployment.