Generative AI Copyright Disclosure Act

Training data transparency

Must AI developers disclose what they trained on?

Detailed disclosure required

Require notice to the Copyright Office of works used to train generative AI

What it says

Anyone creating or altering a training dataset for a generative AI model would have to file a notice with the Register of Copyrights summarising the copyrighted works used (with a URL for public datasets), with civil penalties for non-compliance.

Another US transparency proposal, not enacted.

Sources