Where is the model deployed in Azure OpenAI and how to guarantee the security & privacy?

Question

When I deploy a model, i.e. ChatGPT in Azure OpenAI service, I didn't setup my own compute instance. My understanding is that Azure would help me deploy the model somewhere, but where is this model deployed? Does it share the compute instance (VM) with other people? How do we guarantee the privacy and security over here?

Answer 1

@Kejie Zhao Once a model deployment (consisting of a customer's fine-tuned model or a base model endpoint) is provisioned in a customer's Azure OpenAI resource, data is stored in Azure Storage, encrypted at rest by Microsoft Managed keys, within the same region as the resource and logically isolated with their Azure subscription and API Credentials. The same applies to compute that is used by your deployment. Currently, model deployments do not have an option to use dedicated compute for deployments.

To understand more on the data privacy concerns, please check this page for details on how the data is handled by Microsoft and steps that can be taken to prevent any misuse.

If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.