Governor Gavin Newsom of California recently killed SB1047, a first-of-its-kind artificial intelligence safety bill, arguing that its focus on only the largest AI models leaves out smaller ones that can also be risky. Instead, he says, we should pass comprehensive regulations on the technology.
If this doesn’t sound quite right to you, you’re not alone.
Despite claims by prominent opponents of the bill that “literally no one wants this”, SB1047 was popular – really popular. It passed the California legislature with an average of two-thirds of each chamber voting in favor. Six statewide polls that presented pro and con arguments for the bill show strong majorities in support, which rose over time. A September national poll found 80% of Americans thought Newsom should sign the bill. It was also endorsed by the two most-cited AI researchers alive, along with more than 110 current and former staff of the top-five AI companies.
The core of SB1047 would have established liability for creators of AI models in the event they cause a catastrophe and the developer didn’t take appropriate safety measures.
These provisions received support from at least 80% of California voters in an August poll.
So how do we make sense of this divide?
The aforementioned surveys were all commissioned or conducted by SB1047-sympathetic groups, prompting opponents to dismiss them as biased.
But even when a bill-sympathetic polling shop collaborated with an opponent to test “con” arguments in September, 62% of Californians were in favor.
Moreover, these results don’t surprise me at all. I’m writing a book on the economics and politics of AI and have analyzed years of nationwide polling on the topic. The findings are pretty consistent: people worry about risks from AI, favor regulations, and don’t trust companies to police themselves. Incredibly, these findings tend to hold true for both Republicans and Democrats.
So why would Newsom buck the popular bill?
Well, the bill was fiercely resisted by most of the AI industry, including Google, Meta and OpenAI. The US has let the industry self-regulate, and these companies desperately don’t want that to change – whatever sounds their leaders make to the contrary.
AI investors such as the venture fund Andreessen Horowitz, also known as a16z, mounted a smear campaign against the bill, saying anything they thought would kill the bill and hiring lobbyists with close ties to Newsom.
AI “godmother” and Stanford professor Fei-Fei Li parroted Andreessen Horowitz’s misleading talking points about the bill in the pages of Fortune – never disclosing that she runs a billion-dollar AI startup backed by the firm.
Then, eight congressional Democrats from California asked Newsom for a veto in an open letter, which was first published by an Andreessen Horowitz partner.
The top three names on the congressional letter – Zoe Lofgren, Anna Eshoo, and Ro Khanna – have collectively taken more than $4m in political contributions from the industry, accounting for nearly half of their lifetime top-20 contributors. Google was their biggest donor by far, with nearly $1m in total.
The death knell probably came from the former House speaker Nancy Pelosi, who published her own statement against the bill, citing the congressional letter and Li’s Fortune op-ed.
In 2021, reporters discovered that Lofgren’s daughter is a lawyer for Google, which prompted a watchdog to ask Pelosi to negotiate her recusal from antitrust oversight roles.
Who came to Lofgren’s defense? Eshoo and Khanna.
Three years later, Lofgren remains in these roles, which have helped her block efforts to rein in big tech – against the will of even her Silicon Valley constituents.
Pelosi’s 2023 financial disclosure shows that her husband owned between $16m and $80m in stocks and options in Amazon, Google, Microsoft and Nvidia.
When I asked if these investments pose a conflict of interest, Pelosi’s spokesperson replied: “Speaker Pelosi does not own any stocks, and she has no prior knowledge or subsequent involvement in any transactions.”
SB1047’s primary author, California state senator Scott Wiener, is widely expected to run for Pelosi’s congressional seat upon her retirement. His likely opponent? Christine Pelosi, the former speaker’s daughter, fueling speculation that Pelosi may be trying to clear the field.
In Silicon Valley, AI is the hot thing and a perceived ticket to fortune and power. In Congress, AI is something to regulate … later, so as to not upset one of the wealthiest industries in the country.
But the reality on the ground is that AI is more a source of fear and resentment. California’s state legislators, who are more down-to-earth than high-flying national Democrats, appear to be genuinely reflecting – or even moderating – the will of their constituents.
Sunny Gandhi of the youth tech advocacy group Encode Justice, which co-sponsored the bill, told me: “When you tell the average person that tech giants are creating the most powerful tools in human history but resist simple measures to prevent catastrophic harm, their reaction isn’t just disbelief – it’s outrage. This isn’t just a policy disagreement; it’s a moral chasm between Silicon Valley and Main Street.”
Newsom just told us which of these he values more.
Garrison Lovely (@GarrisonLovely) is a journalist who has contributed to The Nation, Jacobin, the New York Times, BBC Future, The Verge, Time, Vox, and elsewhere. He writes “The Obsolete Newsletter” and is the author of a forthcoming book on the economics and geopolitics of the race to build machine superintelligence
