Is the excellent OCR that lets me select text within an image Apple Intelligence? It uses local models. Or the prediction engine that optimizes battery life by backgrounding apps you're less likely to use, or applies updates while you're sleeping? What about the Latent Semantic Mapping text classification framework, which uses language models for fuzzy document search -- and has been around since 2007?
It wouldn't have made sense in the past to add a master kill switch for all these features, and it doesn't really make sense now to have one for the latest iteration.