Industry & Advocacy News
November 2, 2023
As part of the Authors Guild’s ongoing efforts to protect writers from unfair exploitation by generative AI developers, we submitted extensive comments to the U.S. Copyright Office in response to its call for public input on AI and copyright policy. The Copyright Office will use this information to “analyze the current state of the law, identify unresolved issues, and evaluate potential areas for congressional action.”
As the largest organization representing professional authors in the U.S., we aim to ensure authors’ rights are protected as generative AI technologies transform the landscape for writers.
In our comments, we expressed grave concerns over the threats posed by AI systems like ChatGPT that are trained on massive datasets of books and other copyrighted content without permission or compensation. These AI chatbots can rapidly generate content that imitates human writing, potentially replacing authors and flooding the marketplace with AI-generated works.
“Generative AI poses a serious risk to our members’ professional creative futures,” we stated, citing authors’ fears that publishers will soon use AI to churn out books. Our recent survey found that 70 percent of authors believe publishers will replace human writers with AI-generated books.
We called for new copyright and other laws to mitigate the “profound financial and cultural harm” posed by unchecked AI systems. We argued that consent and compensation must be required for all AI training on copyrighted works.
We proposed that collective licensing through author organizations would offer a path for authors to license works to AI developers and receive royalties for their use, requesting an exemption from antitrust collusion laws for illegal price-fixing risks that could expose writers to expensive lawsuits for jointly negotiating financial terms. We also asserted that extended collective licensing could address past unauthorized uses.
We also emphasized the need for transparency around what training data AI companies use to train their models so that authors and other copyright owners know when their works are used. Public disclosure of copyrighted materials in AI datasets should be mandatory. Further, stricter rules against stripping metadata or other information about the copyright and copyright owner(s) during training should be enacted.
New economic rights that protect creators’ recognizable styles and personas are also needed. Copyright law only covers direct copying on a work by work basis, but AI models can closely mimic authors’ signature voices and bodies of work, and can currently do so without infringing.
Further, conspicuously labeling AI content as such would help mitigate marketplace confusion and deception. We endorsed proposed labeling legislation to achieve this.
Critically, we stressed that granting copyright protection for AI outputs must be rejected. Incentivizing AI-generated works would only hasten the demise of professional human authorship that copyright aims to promote.
Our detailed submission reflects authors’ growing apprehension about AI’s inroads on their careers. By advocating for authors, we hope to spur responsible AI policies that don’t devalue creators’ skills and voices. The stakes couldn’t be higher, as entire creative livelihoods hang in the balance. As we said in our commentary, “We believe that it is inherently unfair to…incorporate the works of creators in the fabric of AI technologies and their outputs without the creators’ consent, compensation, or credit.”
With authors facing declining incomes, loss of freelance work, and other tech disruptions, even small additional threats could devastate the profession. Steps must be taken to prevent this dire outcome. We urge lawmakers to enact safeguards so all authors can thrive alongside advancing technology.
Read our extensive comments here (PDF).
HarperCollins AI Licensing Deal
November 19, 2024
Sign the Statement on AI Training
October 25, 2024
Authors Guild Encouraged by Penguin Random House’s New AI Restrictions
October 23, 2024
