The “generating” from already existing has a particle of a relatively explainable sense in it, unfortunately… to “de-copyright” it…

Simply put, AI has made it easier than ever to launder image copyright infringement… That, in turn, is where AI is changing the game. Previously, content laundering was about fooling computers. Now, it may be able to fool humans.

Source

-–

These two challenges converge on a single point: the need for transparency about what data was used to train a model. Without such transparency…

Source

-–

The foundational principle is deceptively simple: copyright requires a human author… Writing a detailed prompt is not enough.

The machines are producing at extraordinary speed. The law is moving slowly, but it is moving. The businesses that fare best will be those that understood the difference in time.

Source

Even though the models are based on “fair use” of now unknown artists and people, now scrambled and reconstructed… which is all utterly awfully disrespectfully disgusting… indeed…

Also known as “AI-laundering”…

1. Create or fund a non-profit entity to create the datasets for you. The non-profit, research-oriented nature of these entities allows them to use copyrighted material more easily.
2. Then use this dataset to create commercial products, without offering any compensation for the use of copyrighted material.

Source (How Stability AI managed to get millions of copyrighted artworks without paying artists…)

Related: https://en.wikipedia.org/wiki/AI_washing