It isn't clear to me that closed source will matter either given how models are training on copied shit anyway. Reverse engineering will get easier. It's probably goofy easy when a tech company owns larger stacks (like Microsoft owning an OS, Dev tools, server, AI, git repo storage, etc).

It's a weird time. I do hate to see OS getting abused though. There indeed used to be some give and take. But honestly that's been dying in many areas lately.