AB 412: Generative artificial intelligence: training data: copyrighted materials.

This bill requires developers of generative artificial intelligence models to document any copyrighted materials used to train their models. The bill also mandates that developers make available a mechanism on their websites to allow rights owners to submit requests for information about the materials used to train the model. If a request is made, the developer must assess whether the requested material is likely to be present in their dataset and provide a list of materials that match the request within 30 days. Failure to comply can result in discrete violations, and rights owners can bring civil actions against the developer for specified relief. The bill excludes models that are trained exclusively using publicly available data at no cost.