The NWU was invited to testify at the first of a series of listening sessions on the implications of generative artificial intelligence. This is the first step in a lengthy public consultation process by the U.S. Copyright Office, which is hosting the events.
The April 19 session focused on “literary works,” which in copyright jargon includes fiction, non-fiction, advertising and marketing copy, prose, poetry, software code, and other writing. Subsequent sessions in May will focus on visual art (graphics and still photos), audiovisual works, and music and sound recordings. Members of the NWU’s digital media division, the Freelance Solidarity Project, have applied to testify at those sessions. A task force of NWU and FSP members is working to prepare comments to submit to the Copyright Office.
The director of the Copyright Office said they will seek input from the public “in the coming months.” In the interim, we welcome NWU and FSP members who are interested in being part of this and our other copyright advocacy work.
The NWU was represented in this first listening session, appropriately enough, by Edward Hasbrouck, an NWU member whose birth certificate identifies him as the child of a computer program. The NWU’s opening statement was as follows (click here for video):
My name is Edward Hasbrouck. I’m a freelance and self-published independent journalist, book author, and Web content creator. I’m speaking as a member and volunteer for the National Writers Union, which includes writers in all genres and media. The NWU’s digital media division, the Freelance Solidarity Project, who you may hear from at later sessions in this series, also includes creators of digital graphics, audio, and video.
Our members have created works which have been scraped from the Internet, copied, and used for training generative AI without permission or payment and without respect for our moral rights.
The NWU sees (1) moral rights, (2) the right to organizing and collective bargaining for freelance and self-published creators, and (3) feasible and affordable registration of Web content, as prerequisites for protection of our rights as creators of works used to train generative AI.
Training of AI language models begins with copying which, we believe, infringes our copyrights and has already deprived us of hundreds of millions of dollars in rightful revenues. The additional violation of our moral right of attribution makes it impossible for us to tell which of our works have been copied to train AI, and thus frustrates redress for either the economic infringement or the violation of our moral right to object to use of our work to train AI to generate prejudicial content.
Even if copying of our work to train AI is “fair use”, we have the moral rights to attribution and to object to prejudicial use of our work. Congress need not wait for courts to resolve any doubt as to whether copying for AI training is “fair use” to create a means of redress for the massive, ongoing, violations of our moral rights. Generative AI reinforces the urgent need and treaty obligation for Congress to enact effective protection for our moral rights.
As for our economic rights, payments to authors are likely to require collective licensing. But the ability of millions of freelance and self-published creators whose work is used to train AI to bargain collectively with billion-dollar AI companies depends on our ability to organize and act collectively, which is significantly deterred by fear that organizing by freelancers and self publishers might be held to be an antitrust violation.
Congress could best facilitate organizing, collective bargaining, and collective licensing for AI training by explicitly clarifying the right of freelancers and self-publishers to organize and act collectively as workers, including but not limited to collective bargaining over the terms of collective licenses. We should not have to fear that we will be accused of violating antitrust laws if we seek to organize and act together to exercise our rights as writers and digital media workers.
(We’ll add links to this post as soon as the Copyright Office posts the archived video and transcript.)
Several of the other witnesses explicitly endorsed some or all of the NWU’s recommendations. We especially thank Mary Rasenberger, Executive Director of the Authors Guild, which represents book authors, for endorsing our analysis and policy recommendations.
There was particularly broad support, even from witnesses who think that AI developers should be able to copy our work without permission or payment to train AI language models, for our moral right to “attribution” as authors. This means, we think, that works generated by AI should be identified as such and linked to a list of the works used to train the language model from which they are derived.
Other witnesses supported our call for legislation to enable collective licensing by clarifying that it doesn’t violate antitrust law for freelancers and self-publishers to organize and bargain collectively.
And news publishers, who find it as difficult as individual authors to register copyright in their Web-only content, renewed their support for our call for a workable and affordable registration procedure for large and/or dynamic websites.
The U.S. Copyright Office has an unusual dual role as part of both the executive and legislative branches of the Federal government. As an administrative agency, it can, on its own, promulgate regulations and guidance interpreting and implementing the current U.S. Copyright Act. As a division of the Library of Congress, the Copyright Office can advise Congress and make recommendations for legislation to Revise the Copyright Act.
The NWU’s recommendations include regulatory actions the Copyright Office could take on its own, such as issuing regulations to make it easier to register Web content which is being scraped to train generative AI systems but which is now prohibitively expensive and burdensome to register. Our recommendations also include actions that would require Congressional action, including enactment of legislation to protect and allow authors to enforce our moral rights and to clarify our right to organize and bargain collectively without fear of being held to have violated antitrust law.
The most extreme opposition to the rights of the creators of written works used to train AI was voiced by Sy Damle, former General Counsel of the U.S. Copyright Office and now — in typical revolving-door fashion — a lobbyist for Andreessen Horowitz (A16Z). A16Z, whose motto is “Software Is Eating the World”, is the Silicon Valley venture capital firm that led the most recent $150 million round of funding for OpenAI, the company that developed and is commercializing ChatGPT.
Damle argued that AI developers should be allowed to use other people’s written work to train AI without permission or payment, but that those same AI developers ought to be able to register copyright in, and control and profit from, the output of generative AI.
This amounts to saying, in effect, that our work as human creators has no value and should have no legal protection, but that work generated from our work by robots should be valued and fully protected.
Damle also revealed both a lack of imagination and a lack of empathy with authors when he said, as though he thought it self-evident, that he “couldn’t imagine” anyone would think it prejudicial to their reputation to have generative AI, trained on their work without their permission, used to generate works imitating their style that would compete with their own real works.
This is the sort of thinking that will drive debate and government policy on generative AI unless working writers and other digital media workers speak up for ourselves as impacted individuals. Arguments like this on behalf of billionaires and speculators must not go unchallenged. Damle and some other witnesses tried to portray any harm to working authors as “speculation” about the future. But our work is already being used to generate hundreds of millions of dollars for AI companies as well as to train AI to generate fake news, fascist propaganda, phishing spam, and defamation.
We are grateful to the other creators who joined us in testifying to the Copyright Office at this first session on generative AI. We look forward to working with our fellow FSP/NWU members and with our allies to advocate for generative AI policies that respect our economic and moral rights.