California’s AB 412 AI Copyright Bill Faces Criticism as Unworkable

California lawmakers are advancing legislation that would require AI developers to disclose copyrighted works used in training models, but experts argue the bill is impractical and could stifle innovation. The measure, Assembly Bill 412 (AB 412), has drawn opposition from digital rights groups like the Electronic Frontier Foundation (EFF), which calls it “unworkable” due to the lack of a centralized, machine-readable copyright database.

Why is AB 412 Considered Unworkable?

AB 412 mandates that AI developers maintain records of all registered copyrighted works used in training data. However, the U.S. Copyright Office does not provide a machine-readable list of copyrighted materials, and many works are registered without public samples, such as proprietary software code. According to the EFF, this creates a “practical impossibility” for developers to comply, as online content often lacks clear copyright metadata.

“The bill asks developers to cross-reference vast datasets against a system not designed for this task,” said EFF Senior Staff Attorney Matt Scherer. “It’s a compliance burden that disproportionately impacts smaller entities and independent creators.”

What Are the Implications for AI Developers?

The bill’s broad definition of “developer” includes independent creators, open-source projects, and nonprofits, not just large tech companies. While recent amendments exempt universities and government entities, many non-commercial developers remain subject to the requirements. This could deter startups and researchers, who lack the resources to navigate complex compliance processes.

“Small developers can’t afford legal teams to track down every copyright holder,” said Dr. Kate Crawford, a researcher at the AI Now Institute. “This risks consolidating power among established firms that can absorb the costs.”

How Do Courts Address AI and Copyright Issues?

Proponents argue AB 412 would empower creators, but courts have already addressed similar concerns. Federal lawsuits, including cases involving AI training data, have seen mixed rulings on fair use. Some courts have ruled that AI training may qualify as “transformative use,” while others continue evaluating the issue. The U.S. Copyright Office has not issued binding guidelines on AI-specific fair use.

“Rights holders already have legal tools under federal law,” said Professor Rebecca Tushnet, a copyright scholar at Georgetown Law. “State-level mandates like AB 412 risk creating conflicting rules that confuse developers.”

What’s Next for AB 412?

The bill remains under consideration in the California Senate Privacy Committee. Critics warn that without federal clarity, state laws could fragment the regulatory landscape. California has historically balanced tech innovation with creative rights, but AB 412’s critics argue it fails to strike that balance.

California residents can contact their representatives to voice opinions on the bill through the state’s official website. The final outcome could shape the future of AI regulation in the U.S.