For months those of us who follow fair use closely and understand how the doctrine is applied in courts have had to swallow hard and count to 10 while partisans in the AI wars declared, based on no evidence, that all AI training is obviously infringing. These declarations were so obnoxious to the fair use community because we knew, based on hard-fought experience in cases like Google Books, that AI training is about as transformative as a use can be, and that transformative use is generally the driving force in judges’ fair use calculus. Recently, we got a taste of vindication in two opinions from the Northern District of California, which differed in interesting ways but presented a united front on two core elements of the fair use analysis: AI training is highly transformative and licensing for AI training is not a market that copyright entitles rightsholders to exploit. You can read more about them in the links below. It’s early days, but I, for one, felt seen and validated by both judges on these key points.

In darker copyright news, the trend of consolidation and financialization in the music industry took another slouching lurch toward stagnation with the announcement that Bain Capital was partnering with Warner Music Group to launch a $1.2 billion initiative to acquire legacy artist copyrights. While the announcement comes with plenty of fine words about “safeguarding legacies,” the economics of these deals are clear: the catalogs must be exploited to the greatest extent possible in order to make back the huge sticker prices. We can expect more stupid lawsuits like the one brought against Miley Cyrus over “Flowers,” and more pressure to force licensing on fair uses, including machine uses like AI training. Judge Chhabria in the Kadrey v. Meta case and the U.S. Copyright Office in its third report both worried that AI trained on creative works would enable “market dilution” – i.e., competition from AI-powered works that would crowd out new artists. You can be sure that, for the right price, Bain and WMG would gladly license a tool to generate infinitely many new songs based on their “legendary” catalog, with exactly the same effect on new artists. 

Indeed, legacy IP has been crowding out new artists for years—ask any filmmaker not making superhero movies or sequels. As cultural critic Ted Gioia (among many others) has observed, “In the 21st century, creative stagnation is aggressively promoted by entertainment and culture businesses.” If we really wanted a copyright system designed to benefit the next generation, we’d shorten copyright’s duration and explore other steps to help reduce the commercial dominance of decrepit IP assets.