Last Updated on February 26, 2026, 2:16 pm ET
The bipartisan Copyright Labeling and Ethical AI Reporting (CLEAR Act) would require anyone using a training dataset in connection with the training or release of a generative AI model to submit a detailed summary of each copyrighted work in the training dataset to the register of copyrights. The notice must be submitted 30 days before a new model’s commercial use or lease, while existing AI platforms have 30 days from the date regulations are issued; those who do not comply can face lawsuits and civil penalties up to $2,500,000. The register of copyrights would be required to establish and maintain a public database with each notice.
The bill’s broad definitions of “generative AI model” and “copyrighted works” will almost certainly cover libraries training and releasing generative AI models for research and educational purposes. The requirement to report all copyrighted works is unworkable given the scale and scope of library collections, and of generative AI tools. And, the assignment of individual liability could have negative implications for the profession of librarianship.
Notably absent from the bill is any exception for research or educational uses—an omission that disrupts the balance copyright law has long maintained through fair use, and that could expose libraries and academic institutions to liability for activities that have historically served the public interest. But even with carve-outs that align with the fair use exception for libraries and archives, the bill would be unworkable for libraries and archives. For instance, questions around registration status of cultural heritage materials that might still be under copyright risk chilling use by libraries and archives.
While this seems dire, it is unlikely that the bill will become law this congressional session. Even if Congress turns its focus to standalone IP legislation, it is likely to be subsumed by negotiations on higher-priority must-pass legislation. Moreover, the administration is almost certain to oppose legislation that is burdensome to AI development.
The Re:Create Coalition, of which ARL is a member, has opposed the CLEAR Act in a statement explaining how the bill would erase the fair use right.
If you or others in your institution are having conversations with the offices of Senator Schiff (D-CA) or Senator Curtis (R-Utah), consider informing them that research libraries are developers and users of generative AI models, and this bill would impede AI research and development that serves the public interest.
