But then you look at image gen. The established one, namely Adobe, are surprisingly not winning the AI race.
Then you look at code gen. The established IDEs are doing even worse.
I don't rule out the possibility of music being truly special, but the idea of "established tools can just easily integrate AI right" isn't universally true.
What Adobe and others ought to be doing is setting up internal labs that have free reign to explore whatever ideas they want, with no barriers or formality. I doubt any of them will do that.
I'd argue music generation is different from image or code generation. It's closer to being purely art. Take image generation for example. Most of the disruption is coming from competition with graphic design, marketing, creative/production processes, etc. The art world isn't up in arms about AI "art" competing with human art.