Californian AI Copyright Transparency Act and Creative Compensation

Subomi Otunola

California Assemblymember Rebecca Bauer-Kahan has introduced AB 412, the AI Copyright Transparency Act, intended to increase transparency around the use of copyrighted materials in training generative artificial intelligence (GenAI) models. It is common for AI developers to rely on copyrighted material to train GenAI models without acknowledging the owners through credit or compensation. This act follows numerous lawsuits against major AI developers like OpenAI, Microsoft, Meta, and Anthropic for widespread copyright infringement involving books, movies, recordings, and other media.

This act follows the California AI Transparency Act, signed into law last year by Governor Gavin Newsom, which requires generative AI providers to offer an AI detection tool and allow users to include a manifest disclosure indicating AI-generated content. Taking effect on January 1, 2026, California joins Colorado, Utah, and Illinois in requiring AI transparency. However, California is the first to mandate specific watermarking requirements. Bauer-Kahan’s new act expands on this by requiring AI developers to provide copyright owners with detailed information about when their content is used in GenAI training datasets. While the bill has gained support from artists across industries, it has also faced pushback from AI developers and legal experts concerned about the feasibility of tracking the vast amounts of data AI pulls from the internet.

The act addresses long-standing complaints from artists about competing against AI that generates work based on their protected content without verification. A key requirement is the implementation of a Training Data Verification Request (TDVR) tool, allowing copyright holders to check if their intellectual property was used in AI training. The Transparency Coalition has called for the adoption of TDVRs alongside Training Data Deletion Request (TDDR) tools, which would require companies to maintain detailed records of copyrighted materials in their databases. Under this act, creators could request a list of their content used in AI training or confirmation that it has not been included. AI firms failing to comply with these requests could face lawsuits, with a minimum fine of $1,000 per violation. Some advocates are also pushing for a right to request the deletion of their works from training datasets entirely. Currently, only two limited search tools exist for artists and authors, leaving copyright holders largely unaware of whether their materials have been used. Bauer-Kahan stated that "the AI Copyright Transparency Act increases accountability for AI developers and empowers copyright owners to exercise their rights."

However, the tech industry and legal experts question the bill’s feasibility. The current California AI Transparency Act only mandates a "high-level summary" of whether datasets include copyrighted, trademarked, or patented material. Conducting thorough copyright checks on AI training datasets would be a significant challenge, and AI companies fear increased liability due to potentially incomplete copyright lists. The bill also imposes a one-week deadline for developers to respond to copyright claims, creating a burden for major AI firms that would need to process potentially hundreds of requests weekly. Peter Leroe-Muñoz, senior vice president of technology and innovation for the Silicon Valley Leadership Group, argues that this bill could make it "virtually unworkable to train AI on the open web."

The proposal ultimately sets up a lobbying battle between the state’s influential Hollywood  creatives and powerful Silicon Valley tech giants – many of which have much to lose. Courts  move slowly, while AI developers move at breakneck speed and continue to rely on  copyrighted materials to train their AI models which they have built billion-dollar companies  from, while continuing to ignore uncompensated artists. As this is the precedent set by AI  companies, artists from newspapers to Hollywood will continue to wage wars against the  billion-dollar industry’s lack of accountability and compensation. While there is a looming  threat of floods of litigation as well as the arduous work to catalogue copyrighted materials  correctly in the first place – as this bill assumes that there are problem free copyright  databases for all materials – it is ultimately a step in the right direction. This bill reflects the  continuous struggle overcompensation.