More than 150 entities and individuals weighed in against a first-in-the-nation California bill to regulate the most powerful artificial intelligence models of the future.
They included AI companies, business trade groups, venture capital firms, members of Congress and even San Francisco Mayor London Breed.
Opponents outnumbered supporters by more than 3 to 1, according to a tally included in a Senate analysis of the measure.
“I was astonished at sort of the way in which the AI hype machine transcended towards this bill,” said Samantha Gordon, chief program officer at TechEquity, a Bay Area nonprofit that advocates for tech industry regulations and standards.
The piling on of opposition failed to kill Sen. Scott Wiener’s (D) Safe and Secure Innovation for Frontier Artificial Intelligence Models Act. The Senate gave the bill final approval on Aug. 29. But it signaled a new chapter in tech industry lobbying in Sacramento.
After years of fights over data privacy, children’s online safety and social media, AI is quickly becoming the new tech regulation battlefront as states including California seek to place initial guardrails on what could soon become a trillion-dollar industry.
The stakes are especially high in California, the world’s fifth largest economy, home to the headquarters of 35 of the top 50 AI companies, and a leader among the states in tech policy.
“They need to get their positions heard in Sacramento, because what Sacramento does will shape so many other states,” said Thad Kousser, a political science professor at the University of California San Diego.
The push in California and elsewhere to corral AI has put the regulation-averse tech industry on a collision course with state policymakers who say they waited too long to regulate social media and digital privacy.
It has also exposed philosophical rifts over tech regulation between state lawmakers and tech-friendly governors. Colorado Gov. Jared Polis (D) reluctantly signed a comprehensive AI regulation bill this year, while Connecticut Gov. Ned Lamont (D) effectively killed a similar bill by threatening to veto it.
California Gov. Gavin Newsom (D), another friend of tech, is under growing pressure to veto Wiener’s bill. Adam Kovacevich, founder and CEO of Chamber of Progress, a tech industry coalition, predicts he will do so.
“The governor really has an eye on the overall state economy, and he has been very clear that the tech industry is a crown jewel of California’s economy, and he is not interested in prioritizing attacks on tech,” said Kovacevich, who previously worked at Google.
Wiener’s bill aims to prevent worst-case scenarios, such as an AI system triggering a mass casualty event or crippling critical infrastructure. It would require pre- and post-deployment testing of AI models that cost more than $100 million to train, and require that they have full shutdown capabilities.
The Chamber of Progress, among other detractors, derided the bill as “based on science fiction fantasies.” Wiener and his allies said the legislation is needed to guard against AI’s real-world dangers.
The clash over the frontier model legislation activated the venture capital industry, including the influential Silicon Valley investment firm Andreessen Horowitz. In a June podcast interview, one of the firm’s general partners, Anjney Midha, warned that if the bill was passed “it will set a precedent for other states and have rippling consequences inside and outside of the USA — essentially a huge butterfly effect in regards to the state of innovation.”
Proponents of the bill took note of the venture capital opposition. Landon Klein, director of U.S. policy at the Future of Life Institute, one of the bill’s backers, said the industry was able to activate a network of AI companies. The result, Klein said, was an outsized opposition campaign that belied what polling shows as broad public support for AI safety regulations.
“A big part of it is dollar sign related,” said Klein who previously staffed the Assembly privacy committee. “The VC ethos tends to focus on gambling in favor of major profit at the expense of potential risk … and I think this issue in particular is sort of a microcosm [of that].”
The fight over Wiener’s bill grew more fierce as companies including Google, OpenAI and Meta inveighed against it while insisting they support reasonable regulations. The controversy garnered national and even international media coverage.
There were exceptions to the blowback, including support from some big players in the AI community. After Wiener amended the bill in response to industry feedback, San Francisco AI firm Anthropic sent a letter to Newsom calling the bill “substantially improved, to the point where we believe its benefits likely outweigh its costs.” Elon Musk tepidly endorsed the bill in a post to his social media platform X.
While the legislature cleared Wiener’s bill and several other AI-related measures, another high-profile measure that faced intense lobbying was shelved in the session’s final hours.
Assemblymember Rebecca Bauer-Kahan (D), chair of the privacy committee, pulled her bill — which aimed to protect against AI discrimination — after it was radically scaled back in the Senate. She said she plans to reintroduce the bill again next year.
“This was a huge learning year for everyone, and I think that will inform what next year looks like,” said Teri Olle, director of Economic Security California Action, a cosponsor of the Wiener bill. “The California legislature sees itself as having a very important role to play in the AI sector arena. … I don’t think this was a flash in the pan.”
AI bills were not the only focus of tech lobbyists in Sacramento this year.
Assemblymember Josh Lowenthal (D) sought to pass a first-in-the-nation law to fine social media companies up to $1 million for harms to children, amid growing alarm over teenagers’ mental health.
The bill, backed by Common Sense Media, won unanimous votes as it moved through the legislative process. But it also triggered a swift response from tech trade groups and Meta and Snap, according to lobbying activity reports.
By the time the Senate Appropriations Committee advanced the bill, the fines were lower and the legal threshold for triggering the law higher. The amendments hewed closely to Meta’s recommendations.
Read more: Calif. lawmaker shelves bill to fine social media companies for harms to children
Lowenthal said the changes “completely changed” his bill’s intent. Like Bauer-Kahan, he chose to shelve the bill and try again next year. In a post-session interview, Lowenthal said the bill’s opponents strategically targeted the Senate’s appropriations process.
“It means that the tech lobby is confident in its relationships at very specific pain points in the policy-making process,” Lowenthal said.