In a second round of comments submitted to the U.S. Copyright Office this week, the Association of American Publishers insists that copyright law protects authors, publishers, and creators from the unauthorized appropriation of their works by AI developers and slammed assertions by the tech industry that fair use permits developers to use copyrighted works to train their systems without permission or compensation.
"Copyright owners vigorously contest the assertion by tech companies that copyright industries should cease objecting to the unpermissioned and uncompensated use of protected works to create training datasets for Gen AI systems because—as one tech commenter puts it—Gen AI developers have already invested significantly in this new technology on 'settled expectations' of fair use," the AAP comments state. "It is clear from several submissions in the initial comment round that Gen AI developers gambled on a flawed assumption that wholesale reproduction for Gen AI training is categorically fair use, and that policymakers would give them a free pass in the name of innovation, or even more incredible, hardship. It is deeply ironic that these billion-dollar companies bemoan the financial burden they would face if they were required to pay reasonable license fees to the copyright owners whose works are the very building blocks of Gen AI and whose livelihoods are threatened by the same systems."
In the 15-page document, the AAP offered a blunt characterization of tech companies that, in their initial comments, asserted that "the rights of authors and publishers are an obstacle to innovation," calling such claims "nonsense."
"Copyright is the engine of free expression, responsible for incentivizing invaluable, creative expression and information that is inherently transformational to both people and society and indisputably essential to democracy," the AAP argues in its comments. "Copyright protection also fuels markets. The U.S. copyright industries collectively added more than $1.8 trillion in annual value to U.S. gross domestic product in 2021, while offering a vital chain of financial return to support new authorship. Those that seek to weaken these protections disregard the premise that authorship cannot be taken for granted."
The AAP comments come as part of a sweeping study of AI and copyright undertaken by the Copyright Office. With the latest round of comments set to close at 11:59 p.m. December 6, the office has already received more than 10,000 public comments, including from authors, publishers, and other publishing industry associations.
In a release, the AAP highlighted key points from its submission, including:
- “Rather than working with copyright owners, [AI] companies seek to appropriate literature and other invaluable intellectual property for their own commercial gain, and to bend the law to their will. Government should have no role in bestowing commercial advantages to AI companies at the expense of authors, publishers, and other creators.”
- “The companies that benefit from the commercialization of this technology should be required not only to compensate rights holders for their past ingestion of copyrighted works to train Gen AI systems but also for their ongoing and future use of protected works to train new Gen AI systems or fine-tune their existing products.”
- “Gen AI developers are not struggling “start-ups” that need a boost from the government. They count among their investors some of the largest and most profitable technology companies in the world.”
- "It would be a grave error to repeat the past policy mistakes that allowed technology companies to achieve such an unhealthy, monopoly-like market dominance to the point that governments have struggled to curb their power."
- “The issue of national security is certainly of deep concern to all American citizens, particularly where bad actors may use AI and Gen AI systems to sow misinformation or disinformation that undermine our democratic institutions and create other national security risks."
- “Transparency is an essential requirement. It is in the public interest to know what works of authorship have been ingested and an essential part of seeking proper consent to have such information clearly recorded."
- “[I]n no case does the Copyright Act permit unauthorized access to or acquisition of copyrighted works. Lawful access to authorized sources matters. A human is not permitted to illegally reproduce and download 183,000 copyrighted works (the number of infringing titles estimated in the “Books3” corpus) in order to read or learn from them.”
So far, initial copyright lawsuits filed by creators over the use of their works in AI training have faced setbacks in court. As PW reported, a number of cases have had key claims dismissed, including a suit filed by authors.
The complete text of AAP’s reply comments can be found here.