>all AI companies would have to seek out ways to completely retrain their models to avoid copyright infringement, put in place cumbersome methods of restricting

>all AI companies would have to seek out ways to completely retrain their models to avoid copyright infringement, put in place cumbersome methods of restricting the generation of infringing material, or set out to retroactively license all the material they fed into their LLMs. A loss could completely alter the future path of AI (as well as how IP holders are and will be compensated for their work). This is phrased like a negative but all parts of it sound like positive wins for society...