Privacy, Vertex and Claude 3

[Anton231]

Hi everyone.

I am planning to utilize the Anthropic Claude 3 Haiku AI model through the Google Cloud Vertex API. I wanted to understand which privacy and data protection requirements would apply in this scenario — would it be governed by Anthropic's policies as the developer of the Claude model, or by Google's policies for the Vertex platform?

Additionally, I would like to know:

1. Does the Vertex AI platform share any data with the provider of the model, in this case Anthropic? If so, what are the data retention periods for such shared data?
2. In this usage scenario, who is considered the data processor — Google as the provider of the Vertex API, or Anthropic as the provider of the Claude 3 Haiku model?

Thanks in advance

[ruthseki · Accepted solution]

Hi @Anton231,

Welcome to Google Cloud Community!

When you go to Google Cloud Console, navigate to Vertex AI. Select Model Garden>Foundation Models>Claude 3 Haiku(third-party model) and try to click Enable, you’ll see an enablement form same as image below:

This third-party model, licensed from Anthropic and is provided under Anthropic's terms of service, is available through Vertex AI. 

To access the model, users must submit a request form, providing their Google account and Google Cloud project ID. Google will handle this information in accordance with their Cloud Privacy Notice. Anthropic reviews access requests, and Google shares relevant form information with Anthropic for approval. For details on Anthropic's data handling, please refer to their Privacy Policy.

With that being said, privacy and data protection requirements would be governed both by Google's policies and Anthropic's policies. 

In addition, following Anthropic’s official documentation:

Trust Center

All production data is encrypted at rest and in transit. We also ensure to retain customer data in as few places as possible and for a short retention timeline.

---

Subprocessor List

---

Anthropic acts as a data processor, such as with respect to personal data we receive, process, store, or host when you use Anthropic's AI services as or through an enterprise customer. These processing activities are subject to Anthropic's policies, procedures, and other agreements, including its Privacy Policy or Data Processing Addendum.

---

I hope the above information is helpful.