Event banner
Event details
Can you describe how multiple tenancy is handled when customers are using the same underlying models for Co-pilot. Since the GPT models themselves require the prompts to be used as input that activates the layers to generate a response, how are these prompts and responses stored with respect to model use? Are they logged? How are these prompts and responses segregated between customers? Is there a dedicated instance of the model for each customer, or is the same model shared?
Ken Ewert
Microsoft
MicrosoftFor M365 Copilot, there is no bleeding of independent prompt/responses between tenant users. Each is an independent processing where prompts and groundings are gathered, submitted and results returned to the user. It is not "per model" based or anything like that. There will be some logging of prompts in some copilot clients in support of conversation continuity, allowing the user to turn the conversation to refine the last response received.