CA AB412

Requires developers of generative AI models to document any copyrighted materials (registered with the U.S. Copyright Office) they know were used to train their models and make reasonable efforts to identify other copyrighted materials used

Mandates developers provide a website mechanism for copyright owners to request information about whether their materials were used in AI training, including submission of fingerprints and registration numbers

Developers must respond within 30 days to rights owner requests, assessing whether copyrighted materials are present in their datasets and providing a list of identified materials

Rights owners who don't receive required information can sue for $1,000 per violation or actual damages (whichever is greater), plus injunctive relief and attorney's fees; each day past the 30-day deadline counts as a separate violation

Exempts AI models trained exclusively on publicly available free data, used only for noncommercial research, not trained on copyrighted materials, or trained only on materials the developer owns