AI Litigation: What Are the Limits on Scanning to Educate AI?

Anthropic is in court again, sued by the music publication industry for copyright infringement by reason of the scanning of extant music to educate Claude as part of the development of generative AI.

The question: is such use “fair use” and thus permitted under applicable case law. Anthropic offers the argument that GAI will be a transformative advance and that it is anomalous to utilize the copyright laws, designed to foster human progress and creativity, to prevent such actions. Outcome is unclear to me, as for example earlier litigation held that authors could not prevent Anthropic from scanning works that Anthropic purchased but could recover monetarily for “pirated” books; that case cost Anthropic a $1.5 billion settlement (which in the old days was a lot of money).

The professed injury to the plaintiff music publishers is purely economic, but allegedly considerable and something the copyright laws are designed to prevent: the market will be flooded by works that will divert income from human creators of the works which have been copied.

Another example of science presenting equities that need to be weighed in light of laws enacted when the cases now presented could not even be imaged.

Comments are closed.