California’s New AI Bill To Require Copyright Disclosure of Training Data

On February 4, 2025, California Assemblywoman Rebecca Bauer-Kahan introduced AB 412, titled the AI Copyright Transparency Act (the “Act”), which is aimed at increasing greater transparency when copyrighted materials are used as training data for Generative AI (“GenAI”) models and systems. If passed, the Act would require developers who use copyrighted materials as part of their training dataset to disclose this use to the copyright owners.

In a press release Assemblywoman Bauer-Kahan stated that “it is critical for content creators to know if and how their work is being used to train advanced models” and that the Act “increases accountability for AI developers and empowers copyright owners to exercise their own rights.” 

The AI Copyright Transparency Act does not focus on the implications of using copyrighted materials as part of the training data (i.e., whether there is infringement, whether fair use applies). Instead, it focuses on disclosure obligations of the developers to ensure that copyright owners are aware when their copyrighted materials are being used. The proposed legislation aims to further this objective through three key provisions:

• Right to Information: requires developers of GenAI models to disclose to copyright owners a list of copyrighted materials used to train their GenAI model, before the system is made publicly available for Californians to use.
• Online Request Mechanism: requires developers of GenAI models to establish a publicly accessible mechanism, such as an online portal, that allows copyright owners to submit requests for information as to whether the GenAI model is trained on copyrighted materials, and obligates the developers to respond to the copyright owner within seven days with a comprehensive list of materials used to train the GenAI system or model for which the copyright owner holds the copyright.
• Civil Enforcement: Allows copyright owners to bring civil actions against developers of AI models who fail to provide the required disclosures, including the greater of actual damages or $1,000 per violation (each day after the seven-day-response period that the developer does not respond is treated as separate violation), injunctive or declaratory relief; and attorney’s costs and fees.

This Act, if passed, would build upon AB 2013, the California Training Data Transparency Act, which Governor Newsom signed into law in September 2024, and required developers who made their GenAI systems or models available to Californias to post high-level, summary information about the sources of their training datasets. The proposed AI Copyright Transparency Act takes a step further and would require much more detailed disclosure. It would also be one of the first legislations that would require developers to install a Training Data Verification Request (TDVR) tool to allow individuals to query the developer to determine if specific works were used to train a specific model. The Transparency Coalition for AI has advocated for TDVR adoption as the industry standard.

Entertainment and Media Industry Unions Voice Support for AB 412

The Act has received support from sponsor Screen Actors Guild - American Federation of Television and Radio Artists (SAG-AFTRA), and co-sponsors Concept Art Association (CAA) and National Association of Voice Actors (NAVA). SAG-AFTRA’s Executive Director has described the Act as “commonsense legislation” and serves “to protect human artistry and ensure A.I. remains accountable to the work of human content creators.” CAA released a statement lauding the bill’s efforts to increase transparency and to ensure copyright owners are informed when their copyrighted works are used to train materials:

“Thousands of illustrators have found the entirety of their works . . .used to train [GenAI] datasets without anyone having acquired consent, or having offered them any kind of credit or compensation. . . . Because there is no way for creators to verify if their works have been used, its only logical for them to assume that their copyrighted materials are contained within in these systems.”

Meanwhile, NAVA’s support of the Act focuses on the bill providing creators with “more control over how their work is used in GenAI.” In recent years, SAG-AFTRA, CAA, and NAVA have been active in advocating for protections in the entertainment industry to ensure safeguards for their members as AI’s use in media and entertainment continues to increase.

Looking Ahead

Because this Act applies to any developer whose GenAI models and systems are “made publicly available for Californians to use,” the Act is agnostic to where the developer is located, and instead focuses on whether any Californians will use the system, meaning that the legislation could have nationwide, and even global, ramifications. This is an important piece of legislation to keep an eye on.

The bill will likely first be considered by the California Assembly in March. As the bill makes it through the legislative process, Crowell will continue to monitor and report on developments.