AI & Copyright: Fair Use Concerns in Training Data
- Copyright Office report on generative AI training is facing criticism for its assessment of fair use.
- The report's conclusions could influence ongoing court cases, including Kadrey v. Meta platforms, where the report has already been cited.
- Generative AI systems require extensive training on human-created works.
The U.S. Copyright Office report on generative AI training is drawing fire for its stance on fair use, potentially impacting innovation. The report’s analysis, criticized for favoring copyright holders, could influence ongoing legal battles, notably cases addressing the critical intersection of AI and copyrighted material. The central question revolves around whether using copyrighted content to train AI constitutes fair use,a determination that could reshape the legal landscape. This report’s conclusions add more questions than answers, especially for smaller AI companies.News Directory 3 is closely monitoring the implications for technological advancement and the balance between creator rights and innovation. Discover what’s next as courts grapple with thes complex issues.
Copyright Office Report on Generative AI Misses the Mark on Fair Use
Updated May 29, 2025
A recent U.S. Copyright Office report on generative AI training is facing criticism for its assessment of fair use. The report, released amid personnel changes, addresses various aspects of AI with differing levels of detail. At issue is whether using copyrighted material to train generative AI constitutes fair use, a point on which the report’s analysis is considered flawed.
The report’s conclusions could influence ongoing court cases, including Kadrey v. Meta platforms, where the report has already been cited. However, the report’s analysis is not binding and should only be followed if persuasive.
Generative AI systems require extensive training on human-created works. Obtaining permission from copyright holders for millions of works is challenging, especially for smaller AI companies. Fair use would eliminate the need for such permission.
The Copyright Office’s analysis is seen as conflating the use of works for training with the use of models to create similar works. It also allegedly misapplies fair use principles and introduces a novel theory of market harm.
The report’s transformative use analysis is also problematic. Highly transformative uses, serving a different purpose than the original work, are generally considered fair use.Using copyrighted works to develop new technologies is typically viewed as transformative.
The Copyright Office’s analysis also considers whether works were obtained in “bad faith,” a factor the Supreme court views skeptically in fair use cases. Rights holders do not have the right to control fair uses.
The report introduces a “market harm” theory, suggesting courts should consider the overall effects of using models to produce similar works. This theory is not supported by precedent and relies on assumptions about consumer behavior.
Prioritizing licensing markets over user rights could expand the power of media companies and stifle innovation. Fair use is essential for technological advancement, providing room for innovation from various technologies.
What’s next
The courts are expected to resolve the pivotal issue of fair use by applying existing law to specific uses and AI technologies, potentially setting precedents for generative AI and copyright law.