A bipartisan bill introduced in the US Senate on March 16 would require AI developers to submit a formal notice to the Register of Copyrights before releasing any new AI model. PetaPixel confirmed the bill’s core requirements, sponsor identities, and retroactive scope.
The Copyright Labeling and Ethical AI Reporting Act, the CLEAR Act, was introduced by Senator Adam Schiff, a Democrat from California, and Senator John Curtis, a Republican from Utah. The notice requirement targets the training data itself: companies would identify the specific copyrighted works used to train a model and file that information with the Copyright Office ahead of any public release.
Two provisions warrant particular attention. First, the bill applies retroactively, meaning companies with models already publicly available would face disclosure obligations, not just future releases. Second, the bill would reportedly require the Copyright Office to maintain a public database of these disclosures, per reporting on the legislation. The enforcement mechanism, civil penalties and potential rights-holder claims, was not available for independent verification in this package and should be treated as reported, not confirmed.
The CLEAR Act is introduced, not enacted. It has no confirmed committee assignment as of this writing. Its bipartisan sponsorship is notable given the current legislative environment, but passage prospects are unassessed.
For legal and compliance teams at AI companies with models in public release: the retroactive scope is the operationally significant element. If this bill advances, existing models, not just future ones, would trigger the disclosure obligation. That’s a different compliance calculus than most forward-looking AI legislation.