I get the concern, but how would you get AI features without actively sending context to 3rd parties? At least the policy is transparent: doing ai stuff will process your code.

But aside from that, Trylle will implement BYOK and even "bring your own endpoint" pretty soon, that means you'll have modern way of doing software without sacrificing your privacy rights to big labs.