It's taken me awhile to come to this conclusion about AI-generated art because the facts weren't readily available, but now that I've done my research, it's pretty clear. AI-generated art is not ethical and is a detriment to artists.
AI Art generators like Midjourney, DALL-E, and the recently popular Lensa, pull from huge libraries of copyrighted reference material to train their AI. It's very difficult to request that your work is removed, so anyone can generate images using "in the style of (specific person)" and then use those images to sell products.
This doesn't just affect big name artists. These image libraries are massive and may be using your photos or artwork without your permission. You can use the website haveibeentrained.com to see if any of your work was used. It has limitations, though, and is only able to check specific data sets used for image generators like Stable Diffusion.
If you think this couldn't possibly affect you, consider that one artist found photos from their private medical records in the LAION dataset, one of the major datasets used to train these AIs. So even if you have no social media presence, it's still possible that your images are being used without your consent.
AI could be useful to artists for rapidly experimenting with ideas, testing out compositions, and generating reference images. It also genuinely could reduce the income opportunities for trained human artists in the long run.
I think AI art could easily become ethical, if the data was instead based on artists opting in for their work to be used. They could then be credited and paid a royalty fee for derivative works that were created using their images.
However, it is clear, in my opinion, that ethics are not a concern of the companies creating these tools. The reason that most of these tools are free is that their primary goal is to collect as much data as possible. And we really don't know the extent of what this data is being used for.