Simply, you can hold an ideological line against AI. No potentially worthwhile use case outweighs the harms of how it’s being deployed or why it is being deployed.
I've come to the conclusion it's fine to just be blanket anti-AI right now. We can have the nuanced discussions about local models, limited usage, etc. later. It's simply not possible to have those discussions right now. The current emergency is AI as fascist movement. We need to fight that.