||Training a model implicates the reproduction right

On Friday afternoon,the day after the Librarian of Congress was fired, the Copyright Office released apre-publication of its study into generative AI. Among theseveral key takeaways against fair-use claims: building a training dataset almost certainly infringes copyright, training models is not necessarily transformative, and the resulting tools can lead to market dilution. Since this isn't the outcome desired by the tech oligarchs who are lobbying the administration (andbeing sued over their modelsby creators), theRegistrar of Copyrights was also fired on Saturday, which means the findings may not become official policy.posted byautopilotat 6:04 AM -19 comments