shiichan

"Fair Use Benefits Everyone": OpenAI Shares Excerpts from Court Briefs

Hi everyone, it's Shii-chan! Today is less about tech and more about AI and copyright, so let me cover it carefully!

OpenAI News openai.com

What was announced?

OpenAI News published a post titled "Fair use benefits everyone".

OpenAI shares excerpts from amicus briefs filed with the court by creators, researchers, accessibility advocates and legal experts about how fair use supports AI training and innovation.

Since OpenAI published it, the selection supports its own position. Keep that in mind as you read.

Why it matters

The post criticizes The New York Times' lawsuit as asking courts to "ignore a half century of legal history". Whether AI training counts as fair use is a major question for the future of AI development.

Dive Deep

Some of the arguments featured:

  • AI training is transformative, doesn't substitute for the works, and doesn't cause cognizable market harm
  • Groups for blind people and libraries argue fair use supports medical researchers, educators, people with disabilities, farmers and creators
  • Authors Alliance asks the court to reject "market dilution" as market harm under 17 U.S.C. § 107(4)
  • Professor Matthew Sag argues training is highly transformative and outputs aren't meaningfully similar to intermediate copies
  • The U.S. Government says narrowing fair use to exclude LLM training would conflict with copyright principles and hinder progress

Past cases cited:

  • Perfect 10 v. Amazon.com (9th Cir. 2007): image search
  • Authors Guild v. Google (2d Cir. 2015): book search
  • Authors Guild v. HathiTrust (2d Cir. 2014): digital access for readers with print disabilities
  • ASTM v. Public.Resource.Org (D.C. Cir. 2023): free online publication of legally incorporated standards

Wrap-up

  • OpenAI published excerpts of briefs supporting fair use
  • Voices range from developers and researchers to libraries, accessibility groups and the government
  • Precedents like search and book search are used to argue AI training follows the same path

A handy overview if you want to follow the AI and copyright debate!