Washington, D.C. – Re:Create issued the below statement today after the U.S. Copyright Office released a prepublication version of its findings on generative AI training and licensing issues:

“While Re:Create appreciates the diligent work the Copyright Office has done on its report on generative AI training, and we agree that no changes to copyright law are needed to accommodate AI innovation, the Office misinterprets some key fair use principles. Undue licensing requirements could cause serious harm to creativity and innovation, and courts should recognize that fair use is a right that persists even when a license is available,” said Re:Create Executive Director Brandon Butler. 

“A developer’s use of copyrighted works to train generative AI tools is typically fair use and should not require licensing. We do hope lawmakers will heed the Office’s advice and avoid any changes to copyright law before the courts decide the fair use implications of AI training.” 

Below are Re:Create’s key takeaways from the report.

• Re:Create agrees that…
  • Existing fair use law is sufficient for evaluating and balancing the copyright interests at stake in AI training. (p.107)
  • The markets for licensing seem to be developing naturally in places where licenses are useful and mutually beneficial, so there’s no sign of a need for the government to get involved to encourage licensing. (p. 107)
  • Training a generative AI foundation model on a large, diverse dataset will often be a transformative use. (p. 45, p. 107)
  • Where the purpose is transformative, use of entire works will be favored by the third fair use factor. (p. 55, p. 57)
  • When AI tools have “guardrails” to prevent infringing/memorized outputs, that favors fair use under the third factor. (p.59)
  • When licensing for AI training is impossible, there is no market to harm and the 4th factor should weigh in favor of fair use. (p. 71)
  • The use of pirated data should not be conclusive against fair use. (p.52)
  • “As generative AI involves a spectrum of uses and impacts, it is not possible to prejudge litigation outcomes.” (p. 74)

• Re:Create disagrees with the report insofar as it…
  • Misapplies key fair use principles, such as the transformative use analysis, by focusing on the audience for an AI model’s outputs rather than whether the model itself is a new and transformative creation.
  • Misunderstands how training can be characterized as “non-expressive use.”
  • Gives weight to “market dilution,” a theory of market harm that the Copyright Office concedes has never been endorsed by any court, and that runs counter to the primary goal of copyright: to encourage the creation of new works. 
  • Prioritizes licensing over user rights, falling into the circular argument that licensing is required whenever a licensor is willing to be paid, a view that would dramatically expand the market power of copyright holders.
  • Penalizes copying of unprotected stylistic elements, a chilling outcome for both the creative and innovative industries.

About Re:Create: Re:Create is a coalition comprised of a broad membership of think tanks, advocacy organizations, libraries, technology companies – large and small – that serves as the leading coalition united in the fight for a balanced copyright system that is pro-innovation, pro-creator, and pro-consumer. Not every member of the Re:Create Coalition necessarily agrees on every issue, but the views we express represent the consensus among the bulk of our membership.