At least if they either use the user's code in training OR if it uses any code with unknown provenance.
Microsoft's solution seems to be the best: we take responsibility for the code. But that responsibility comes at a cost.
And, if you want AI but only using local models, all the same UI API’s are available to you. (modulo some know-how)
JetBrains makes it clear that the code is not submitted for training purposes, contrary to the speculation in a lot of comments.
It’s also not enabled by default and requires a trial or subscription as well, so it’s not something people are going to stumble into accidentally using.
I’m sure some strict companies will forbid the use of any product that might possibly have an AI component, but they’ll be refining policies and making exceptions as everything from operating systems to web browsers start coming with optional AI features.
There are also many other cases where it is just outright prohibited, or heavily restricted, even if the aspects you describe were allowed.
Compliance-related challenges due to potentially biased decision making.
Financial regulation compliance concerns.
Government regulations around the application of AI/ML in certain circumstances, or model use restrictions.