Forum Discussion
Copilot, Sharepoint Online, and LLM Training Data
When one creates a Copilot for a SPO site, what data becomes part of the model's training? Are prompts used by msft in training the model? Is any part of the site data or AI's reponses used in training the model? This is a critical issue for any organization with protected data, like law firms, health care delivery orgs, etc.
- Is Copilot not just doing a smart/generative answers response to an Enterprise Search scoped to the site? I don't believe it is creating its own LLM.
My understanding is that the standard index for SharePoint is not a vector index which is why Copilot Studio has some challenges too.
Microsoft have been pretty strong on the no data used to train the model across their Copilots so I would say Copilot in SharePoint would/should be the same,
