AI Regulation Still in the Mix as California Legislature Returns to Session

California legislators return from summer recess on Monday, and among the bills they’re considering are several that, piece by piece, aim to regulate generative artificial intelligence.

Six such bills go before the crucial Senate Appropriations Committee on Monday, and all are expected to be moved to the “suspense file,” meaning the committee will take up their fates on Aug. 15. Lawmakers are optimistic about their chances; none have inspired the same level of industry backlash as SB 1047 by Sen. Scott Wiener (D-San Francisco), which is headed to the Assembly Appropriations Committee later this week.

“We just need to do the right thing,” said Assemblymember Rebecca Bauer-Kahan (D-Orinda), who chairs the Assembly’s Privacy and Consumer Protection Committee. She is pushing four AI bills this year, including AB 1836, which would prohibit using “digital replicas” of a dead person in an expressive audiovisual work or sound recording without prior consent, and AB 2930, which would prohibit the use of automated decision tools that pose a “reasonable risk” of algorithmic discrimination.

“I think the lawmakers are there. The public is there. This is polling at astronomical rates. Communities want to be protected from the risks of AI. And so we’re going to do it,” Bauer-Kahan said.

A recent poll by the Pew Research Center found that 67% of Americans familiar with chatbots like ChatGPT are more worried that government regulations will not go far enough than that they will go too far.

Bauer-Kahan added that California lawmakers are open to industry input on their bills and that Silicon Valley knows something is likely to hit Gov. Gavin Newsom’s desk before this legislative session is over. It’s just a question of what, exactly.

“I think they understand that the public and lawmakers are not going to go for another go-round of no regulation,” Bauer-Kahan said.

Industry responses to California bills are all over the place

California dominates the Forbes AI 50 list. San Francisco alone is home to 20 of what Forbes considers the most promising privately held AI companies.

For the most part, Silicon Valley lobbying groups have not aired their concerns about the California bills outside of Sacramento — with one notable exception: Wiener’s SB 1047, which has inspired open letters warning gravely that it “could inadvertently threaten the vibrancy of California’s technology economy.” Regardless, that bill is still in the running.

Some of the bills enjoy strong or qualified tech industry group support, including AB 2355, which would require labeling of political ads made with generative AI, and AB 2839, which would prohibit misleading election season communications.

Big-name venture capitalists fighting AI regulation at the state and national levels have claimed they’re worried about the little guys in the AI ecosystem. But AI Tech is not a monolith.

“The AI revolution is already underway, and it’s as likely to be happening at your local bike store as it is with the Microsoft products on your desk,” said Morgan Reed, president of the App Association, which represents the companies that make smartphone apps.

In the absence of federal regulation, Reed said state lawmakers are attempting to flood the field. Something on the order of 500 bills is pending across the country, he estimated, and he said lawmakers of all stripes need to think about the tiny AI tech deployers as well as the big developers like Google, Apple and Meta.

“Tiny companies can have millions of records. And in that case, they’re handling millions of Californians’ data. And they need to take that just as seriously as a really large company,” Reed added.

California is not in the lead vis-a-vis AI legislation

While California has historically taken the global lead in some forms of comprehensive regulation, such as with air pollution and privacy, the same can’t be said when it comes to AI regulation. Other states, like Colorado, and regions, like the European Union, have moved ahead.

“In light of federal inaction, we have a piecemeal approach because each state is moving legislation along separately,” said Stanford professor Daniel Ho, who focuses on the intersection of law, political science and computer science. Even at the state level in California, he said, the “attempts to plug all sorts of different holes … are not quite the kind of comprehensive approach that you’re seeing in something like the EU AI act.”

Colorado, experts say, tried to model its state legislation after the European Union’s AI Act. The state’s new law, which is set to go into effect in 2026, aims to protect the public from bias or discrimination embedded in AI systems and to set guardrails to make sure the technology is used ethically. The law also addresses many of the issues individual bills in California attempt to address one by one.

The philosophical approach underpinning most attempts at regulation

“AI is not a thing like a train or even a railway system, right?” said Ryan Calo, a professor at the University of Washington School of Law who is co-director of the UW Tech Policy Lab. “It’s best understood as a set of techniques that are aimed at approximating some aspect of human or animal cognition using machines. So you can’t regulate AI as such.”

Whether the legislation is comprehensive, like Colorado’s, or piecemeal, like California’s, Calo explained, lawmakers are looking to dial in on how the use of the software impacts consumers and citizens.

“So there may be certain things like emotional manipulation, or trampling on human rights, that are just not allowed,” Calo said, explaining the kind of harm or impact that lawmakers could explicitly prohibit. A second tier of legislation could focus on closely regulating software uses that have the potential to damage consumers, perhaps reducing their access to employment or healthcare.

Finally, Calo said, “There may be a third tier of things like spell check, Spotify playlists that may be subject to very little regulation at all.”