A bill being considered in the California Assembly would require artificial intelligence developers to notify copyright owners such as actors and artists when their materials are used to train an AI model, providing them an opportunity to exercise their rights.
Developers of generative AI models would have to document when they use copyrighted materials and respond to requests from copyright holders who wish to see a list of their materials that were used to train the model.
At a press conference Tuesday ahead of a scheduled hearing on the bill, Assemblymember Rebecca Bauer-Kahan (D), chair of the Privacy and Consumer Protection Committee and the bill’s author, said it was a matter of fairness.
“They have a right to be protected as they create work, as they put it out into the world and as they want to be paid for it,” said Bauer-Kahan, who was flanked by leaders from the entertainment industry.
Bauer-Kahan said the legislation builds upon a law passed last year that requires AI companies to make their training day public and disclose whether any copyrighted, trademarked or patented material was used.
The measure is cosponsored by the Screen Actors Guild-American Federation of Television and Radio Artists, Concept Art Association, National Association of Voice Actors, and Authors Guild.
“My union has no desire to halt technological progress,” SAG-AFTRA Secretary-Treasurer Joely Fisher said at the press conference. “We do however want to make sure that AI improves human lives rather than harming or exploiting us.”
Karla Ortiz, an illustrator with major movie credits who spoke on behalf of the Concept Art Association, said “almost the entirety” of her work was taken without her consent to train for-profit AI models. She subsequently filed a class action lawsuit against several companies.
“It is a technology that uniquely exploits the hard work, creativity and property of others. No other tools are like this,” Ortiz said.
Concerns about AI models mimicking the voice, image and likeness of artists is especially acute in California, home both to the film industry and 32 of the 50 top AI companies in the world.
OpenAI last year introduced a personal assistant voice that sounded like actress Scarlett Johansson. The company, which had reportedly sought out Johansson to lend her voice to its product, denied any connection but quickly abandoned the synthetic voice in question. Bauer-Kahan invoked that incident Tuesday.
“One of the things about AI and these large language models is that what is in them, in a lot of cases, is a mystery … and then out comes Scarlett Johansson’s voice,” Bauer-Kahan said.
Open AI and Google recently made the case to the Trump administration that they should be able to train their models on copyrighted materials under the fair use doctrine. The New York Times, famous authors and others have sued OpenAI alleging copyright infringement. Those lawsuits and similar cases against other AI companies have yet to be resolved.
Bauer-Kahan said her legislation would not change fair use law, which allows some copyrighted material to be used without permission or payment. Instead, she said, her goal is to help artists answer the questions: “Is that my voice? Is that my work? Is that my art?”
The bill would allow artists to sue AI companies that do not honor requests for information about copyrighted material and collect $1,000 per violation. Bauer-Kahan said she planned to amend her bill in committee to allow copyright holders to label or “hash” their works so that it is technically possible for AI companies to comply with their requests for information. Another amendment would exempt generative AI models that make their data available to the public for free.
The bill’s opponents include the California Chamber of Commerce, tech associations and the Electronic Frontier Foundation. Among their arguments, according to a bill analysis: The legislation is premature, preempted by federal law, could reveal trade secrets, and will lead to a litany of lawsuits.
In a statement Tuesday, the Computer & Communications Industry Association said the bill “would cripple AI innovation in California.”
“California should be fostering AI innovation, not passing laws that will push jobs, investment, and groundbreaking research out of the state,” Aodhan Downey, the association’s state policy manager, said in a statement.