I mean, in a legal sense, no -- it's not theft of the proceeds. You may mean that you think it's morally/philosophically similar to that, in the same sense as piracy and other IP violation, but these things are very specifically not classified as theft from a legal standpoint.
Violating the intentions and purposes of the creators is in some cases an IP violation, and in others it isn't -- it all depends on the details. A lot of AI training on IP-protected materials can fit into the bucket that isn't an IP violation (or any other legal violation), though much of the historical actual training data wasn't collected this way.
IP creators don't have an amorphous right to "a return" or to have their intentions and purposes strictly respected. They have very specific rights in law, and those often don't match up 1:1 with expectations, and almost never with a specific set of the creator's intent (except where that aligns with already-existing law).
They aren't making a legal claim, "What they did was technically legal" is almost never a compelling argument in favor of behavior, and the fact that people can't expect legal protections from this behavior is exactly why they're taking matters into their own hands, leading to the information protectionism that started this comment thread.