Military would be fine, because they don’t tend to update very frequently, if at all. If it works, that’s the way it will stay, and the recent controversy wouldn’t exactly encourage them to do so.

What about its use in a company that has extremely valuable trade secrets that need to be kept that way?

Same way the LLM debacle has currently gone, where people will just throw sensitive information into it with abandon. At least one major tech company has penalised workers for doing that with ChatGPT.

If there’s a group policy to turn it off, maybe, but Microsoft might just not have one, or it’ll need to be disabled every update.