I see nothing wrong with creating an AI image, which you then use as a reference for painting. And I think the ethics of it would be similar to using photographic references. Just as if you take your own photo and use it as a reference, so too if you make your own AI image to use as a reference, there should be no impetus to qualify the painting as such. In other words, the painting would be ex-."" 16"x20" Oil on canvas. ""-- and not ""16x20" AI oil on canvas"-- or some variation
In fact technically, any AI image used as reference - self-created or not - would not require permission or to be divulged -- since AI images cannot be copyrighted
Several years ago I made some AI images using an early generation of Dali E. for the purpose of painting references. They were laughably out loud ridiculous - totally unusable. This due to my inability to write a decent prompt and the program being so primative.
Now things have changed. And I plan to try AI again- and I see nothing legally or ethically wrong.
Apologies for this rather late reply - I have been laid low first by food poisoning or something like it, and then by the flu. Now I'm finally slowly returning to life.
Now I don't want to endlessly debate the issue either, because if truth be told, I have not quite decided yet how I feel about it myself.
Seems to me there are several possible lines of argument. Artists often complain about AI stealing their work. But if we do that, and then use AI-generated imagery to paint from, surely we are then also stealing? Also on the con side: some of the pictures generated by AI are actually quite beautiful, and at least on screen look for all the world like real paintings. I worry that if I start working from such images, I'll end up getting very lazy. The AI has already done most of the creative work!
But one can also argue from a different angle. E.g. take Caravaggio and his followers: he came up with a radically new style of painting, and within a decade, Europe was filled with "Caravaggists" all painting in the same style. I.e. they trained themselves on his style and methods and followed suit. And one may ask: isn't this really precisely what AIs do when they get trained on this or that artist's style? We can copyright a picture, but we cannot copyright a style, and thus, I am not too sure that there is in fact anything wrong with using AI to generate whatever pictures we like. It is bad news though for artists such as illustrators who make their income primarily from reproductions. But then, it wouldn't be the first or last time that entire categories of jobs got supplanted by machinery.
Another thought occurred to me: any and all art has always been a combination of a human mind, hand and some or other form of technology. This goes for everything from cave art to Andy Warhol prints. And thus, one may argue that using AI to generate reference imagery is just one more example of using whatever technology is available.
As I note above, I have not quite worked out what my own feelings about it are. The temptation is there, because AI imagery is now ubiquitous - you don't even have to write your own prompts, because perfectly beautiful and usable images are all over such sites as Pinterest, and because they are AI-generated, they are copyright-free. (It's a good question whether, if you make a copy of such an image, you can then copyright your copy!)
AI can generate quite beautiful pictures nowadays. What it cannot do is to turn any of them into original, hand-painted pictures. So maybe we're witnessing a whole new thing in art, where the artist's main job is to translate the AI-generated image into attractive brush strokes. Sooner or later printing tech will catch up, and prints that have a brush stroke effect so perfect that it looks like real human-made brush strokes will become cheap and widely available, and then perhaps the remaining pool of human artists will shrink a bit more.
Anyway, such are my dubious thoughts about the issue.