Bill > AB412

This bill requires developers of generative artificial intelligence (GenAI) models, which are systems that can create new content like text or images based on their training data, to provide a way for owners of copyrighted materials, referred to as "rights owners," to request information about how their "covered materials" (registered works or pre-1972 sound recordings) were used to train these models. Developers must maintain a website mechanism for these requests, document them, and within 30 days, assess if the covered materials were used and provide a list of any identified materials. Failure to comply can result in daily violations, and rights owners who follow the process can bring civil action for damages, injunctive relief, and attorney fees, though developers are given a 30-day window to cure any violations after notice. The bill includes exceptions, such as for models trained exclusively on publicly available, free data, or for models not trained using covered materials, and clarifies it does not impose liability on telecommunications or information service providers acting solely as conduits.