Introduction
The debate around artificial intelligence regulation often centers on a perceived trade-off: protect innovation or protect society. Critics argue that state-level rules could stifle U.S. competitiveness in the global AI race, particularly against China. Yet California’s new AI safety and transparency law, SB 53, signed this week by Governor Gavin Newsom, is emerging as evidence that thoughtful state legislation can achieve both goals. According to Adam Billen, vice president of public policy at youth-led advocacy group Encode AI, the bill demonstrates that regulation doesn’t have to hinder progress—it can reinforce it.
Main Content
At its core, SB 53 requires large AI labs to disclose their safety and security protocols and prove they are taking steps to prevent catastrophic misuse, such as cyberattacks on critical infrastructure or the development of bio-weapons. Enforcement falls under the California Office of Emergency Services. While many AI companies already conduct safety testing and release documentation like model cards, Billen argues the bill is critical to ensure firms don’t cut corners under competitive pressure.
He points to OpenAI’s own stated policy of potentially relaxing safety standards if rivals release risky models without similar safeguards. State laws like SB 53, he says, hold companies accountable to their promises.
Industry opposition to AI regulation has been fierce. Tech giants like Meta, influential VCs such as Andreessen Horowitz, and leaders like OpenAI’s Greg Brockman have invested heavily in political efforts to block state AI laws, including supporting a proposed 10-year moratorium on state regulation. Encode AI, however, helped lead a coalition of more than 200 organizations to strike down that initiative, emphasizing the importance of state involvement in AI oversight.
Meanwhile, federal lawmakers such as Senator Ted Cruz have pushed proposals like the SANDBOX Act, which would allow companies to bypass certain federal AI rules for up to a decade. Billen warns that such federal preemption risks “deleting federalism for the most important technology of our time.”
The geopolitical context also looms large. While Billen agrees the U.S.–China AI race is real, he argues the right levers to ensure U.S. leadership are federal export controls and domestic chip production support—not rolling back state regulations. Legislative efforts such as the CHIPS and Science Act and the proposed Chip Security Act aim to strengthen U.S. control over advanced semiconductor supply chains, but industry resistance has been mixed, especially from Nvidia, which relies heavily on China for revenue.
Against this backdrop, SB 53 represents what Billen calls democracy in action: messy compromises between policymakers, industry, and civil society that still deliver meaningful results.
Conclusion
California’s SB 53 underscores an important point in the AI regulation debate: state laws need not be roadblocks to innovation. Instead, they can provide a framework that enforces safety without smothering progress. While federal preemption battles loom and geopolitical tensions with China complicate the landscape, SB 53 shows that collaboration between lawmakers and industry can produce balanced, forward-looking rules. For Billen, it’s proof that democracy and federalism still work — even when applied to one of the most transformative technologies of our time.