© Kyle T. Webster member since 2020
ASCRL has endorsed AB 412, titled “Generative artificial intelligence: training data: copyrighted materials. The bill was introduced into the California State legislature by California Assembly Member Bauer-Kahan. The bill would require generative AI developers to document both the copyright owners and any copyrighted materials used to train the system or model before the platform is made publicly available. Developers would also be required to provide copyright owners a comprehensive list of the works used for training upon receiving a duly submitted request. Should a developer fail to respond to the request, copyright owners would be authorized to bring a civil action.
In our letter of support, we wrote: “By increasing transparency around the training of generative AI models, AB-412 holds AI companies accountable for the use of copyrighted works in their training datasets. Additionally, by specifically targeting the foundational large-scale training of LLMs, AB-412 will not adversely affect the generative AI startups which largely employ retrieval-augmented generation techniques.”
The bill was co-sponsored by Concept Art Association and the National Association of Voice actors, and is supported by a number of organizations, including the National Writers Union, Music Artists Coalition, Romance Writers of America, Songwriters of North America, Transparency Coalition, SAG-AFTRA, and Writers Guild West.
The full text of the bill can be read here.