Monday 23 June - Federal court decision ruled as “fair” and “transformative” the use of copyrighted books to train AI, in the case of three authors against Anthropic.
Mystery best-seller Andrea Bartz and non-fiction writers Charles Graeber and Kirk Wallace Johnson had filed a lawsuit against Anthropic, an AI firm backed by Alphabet, the parent firm of Amazon and Google. The authors claimed that their work was used without permission for the training of Claude AI model, leading to multi-billion gains for the AI firm.
Judge William Alsup considered Anthropic's training use of the writers’ books as lawful, because it was "transformative" and similar to the use of any aspiring writer, who reads others’ books in order to learn how to write.
Nevertheless, the case was not dismissed as requested by the AI company. The Judge ruled that the firm had created a “central library” of seven million pirated books, in order to train its AI models. Although the firm paid for the books later in the process, the court ruled for a next trial on the grounds of piracy, where Alphabet may have to pay damages of up to $150,000 per copyrighted book.
The Judge noted that the authors did not claim infringement of their rights for the outputs of the AI model, but only for the training. Had they done so, the case would be different. For example, in the recent lawsuit of Disney and Universal against Midjourney, the plaintiffs allegedly accused the AI firm for replicating their copyrighted works in the outputs of its generative models.
The decision is expected to have an important impact as a landmark ruling, relevant to ongoing trials over AI alleged use of copyrighted works.
Find more here
Image by Pete Linforth. Free for use under the Pixabay Content License.