International Publishers Association (IPA) Offers Statement on Generative AI
Information Industry News
Geneva, Switzerland | February 21, 2024
Generative AI companies depend on creative works of authorship to fuel and train their tools and products. Some of these companies claim that they do not have to seek permission or pay for the works they are so freely usurping for their advantage, but this is false.
Works of authorship—which include scores of fiction and nonfiction such as award-winning novels, renowned children’s books, ground-breaking scholarship, biographies, scientific publications, political memoirs, and more—are not free for the taking. Rather, these works are economically incentivized and legally protected by both national copyright laws and international treaties and have been for centuries. This legal protection is neither debatable nor dispensable and there is no good public policy reason to weaken it for the convenience and objectives of technology companies, some of which are the most dominant corporations the world has ever known.
It has become commonplace for some generative AI companies to highlight their innovation and investments as a reason for governments to give them broad copyright exceptions or regulatory loopholes when it comes to existing copyright requirements. But these companies, some of which are valued at more than 80 billion US dollars, owe their considerable success to the prior innovations and investments of others—the
intellectual and creative investments of authors and the financial investments of publishers.
Publishers have always embraced technology and technology partners. They are early adopters of many AI tools and believe fundamentally that human expression and technological innovation are symbiotic. However, the legal framework is clear. Collecting, handling, storing, and copying works of authorship to train AI models implicates the exclusive rights of authors, which cannot be ignored. Generative AI companies must license works of authorship that they do not own—and seek to use for their gain—in the manner directed by the rightsholder.
Basic copyright principles also align with broader ethical considerations for society. They will facilitate transparency for the public, reduce misinformation, and most importantly, ensure that the world will continue to have authors, publishers, and a thriving creative community. Transparency as to which works have been used to train AI models is critical: it protects the public from harm and ensures that authors will not be stripped of attribution for their works.
Licensing is simple in the digital age, and there is no excuse to ignore it. Direct and collective licensing models are ubiquitous, flexible, efficient, and continue to evolve. Technology companies can respect copyright and avoid liability. There is no need to except them from the well-established rules of copyright. It should also go without saying that pirate sites are a scourge on society and must remain off limits as a basis of AI training and profits, not only because the content is illegally appropriated from rightsholders, but also because it is untrustworthy.
Copyright is essential to reading, learning, and independent thought, and to the survival of the global publishing industry. Yet these issues are ultimately about the public interest itself. Because democracies depend on authors and publishers, we call on governments to stand up for copyright and resist the calls to give technology companies more power over the public.
Stand up for society by putting authors, publishers, and readers first.