California passes first-ever AI safety law targeting big tech

California passed an artificial intelligence (AI) Senate Bill 53 (SB 53), which was signed into law last week by Governor Gavin Newsom. The regulation is a first-in-the-nation legal framework that requires large AI companies, specifically those earning more than $500 million annually, to publicly disclose their safety and security protocols. 

Adam Billen, vice president of public policy at youth-led advocacy group Encode AI, stated in a recent podcast interview with TechCrunch that SB 53 demonstrates how government regulation can complement technological advancements.

The legislation mandates these firms to outline how they prevent their systems from being misused in catastrophic ways, such as in cyberattacks or the creation of biological and chemical weapons.

‘Regulation and innovation don’t have to clash’

“The reality is that policymakers know we have to act,” Billen said on the Equity podcast. “There is a way to pass legislation that genuinely protects innovation while ensuring these products are safe.”

Billen argues that many of the bill’s requirements are practices that leading AI firms already follow, including model testing and transparency reports. Still, he noted that competitive pressures have led some companies to relax their safety protocols, which the new law seeks to address.

“Companies are already doing the stuff we ask them to do in this bill,” he said. “Are they starting to skimp in some areas? Yes. And that’s why bills like this are important.”

As seen in a letter sent to Governor Newsom before the bill was passed, OpenAI refuted the then-proposed law, arguing that AI regulation should fall under federal jurisdiction, not individual state governments. 

Venture capital firm Andreessen Horowitz doubled down on OpenAI’s plea, suggesting in a blog post that certain state-level AI laws could violate the US Constitution’s dormant Commerce Clause, which restricts states from interfering with interstate commerce.

Billen dismissed such objections, saying claims that state laws threaten innovation or competitiveness are overstated. 

“Are bills like SB 53 the thing that will stop us from beating China? No. It’s intellectually dishonest to say that’s what will hold America back.”

The federal vs. state AI regulation power struggle

Encode AI, the group Billen represents, has previously led a coalition of more than 200 groups to oppose federal preemption proposals that would block states from enacting their own AI rules.

According to Billen, US Senator Ted Cruz is among policymakers looking to undermine state autonomy in AI oversight. Cruz introduced the SANDBOX Act on September 10, which would see AI companies apply for waivers to temporarily bypass certain federal regulations for up to a decade.

The Encode AI VP expects a forthcoming proposal for a federal AI framework that might appear balanced but, in practice, could override state-level laws. He warned that such legislation could erase federalism in digital America and expose the government to a negligent community where companies can go wrong, without dealing with the consequences.

Safety, federalism, and competition with China

Billen admitted the US-China competition is real, but is not enough to warrant the blocking of state-level efforts in AI regulation. He said lawmakers should focus on federal export controls and help American AI firms access computing chips they need to compete globally.

“If the thing you care about is beating China in the race on AI, and I do care about that, then the things you would push for are export controls and ensuring American companies have the chips,” he continued, “but that’s not what the industry is pushing for.”

Per Billen, SB 53 is a functioning example of democracy in action, a collaboration between government and industry that produced a law both sides could live with, even if imperfectly.

“It’s very ugly and messy. But that process of democracy and federalism is the entire foundation of our country and our economic system. I think SB 53 is one of the best proof points that that can still work,” he concluded.

The smartest crypto minds already read our newsletter. Want in? Join them.