Turn off Copilot at semantic model or report level

Copilot is great, or it will be great some day. But at present the only options are to turn it on for all reports & semantic models in a tenant, or turn it off. Please provide more nuanced control e.g. turn off for an individual report or semantic model. Workspace or capacity might be other convenient control points.

"Security" in the wider sense for Power BI often relies on users with Viewer access only having access to pages with visuals, filters etc that constrain what data they can view. Copilot has the potential to be a new "vector" revealing data that has previously remained hidden. As it is a totally obscure feature to Power BI authors/admins and allows Power BI viewers to type in questions, there doesn't seem to be any practical way to guarantee particular results cannot be extracted.

We already have this capability for Q&A - can turn off at the PBIX file or semantic model level.

Many semantic models are not ready to be exposed via Copilot - there are many preparation steps that may be required e.g. renaming columns, dropping redundant data. Surely it is better to give report authors & workspace admins control over when to reveal a semantic model via Copilot? At present, it is open slather for all semantic models, so resources are being wasted generating Copilot results that are often gibberish or halucinatory. This is a poor first experience for Copilot users, and can damage the brand of that product as well as Power BI.