California's SB 53, an AI safety bill, mandates transparency from major developers on catastrophic risks, pending Newsom's approval.
The bill applies to companies with over $500 million in revenue, requiring them to publish safety frameworks for significant AI risks.
Industry is divided; Anthropic endorses the bill, while OpenAI warns against state-level regulations stifling innovation.
Newsom's decision on SB 53 could establish a major AI safety law, impacting the debate on state versus federal regulation.
California lawmakers passed SB 53, a landmark AI safety bill that now heads to Governor Gavin Newsom's desk. The legislation requires major AI developers to be transparent about the catastrophic risks of their most powerful models, placing the home of Silicon Valley at the center of the AI regulation debate.
Reading the fine print: The bill targets "large frontier developers," a group first defined by Politico as companies with over $500 million in annual revenue. It mandates they publish safety frameworks and defines "catastrophic risk" as any event causing over 50 deaths or $1 billion in damages, including from AI-assisted bioweapon creation or autonomous cyberattacks.
A divided industry: The measure has split the tech world. Anthropic publicly endorsed the bill, arguing it formalizes best practices and creates a level playing field. Conversely, OpenAI and other groups have lobbied against it, warning a "patchwork" of state laws could stifle innovation and calling for a single federal standard instead.
Déjà vu for the governor: This is the second time a major AI bill from State Senator Scott Wiener has reached Newsom. He vetoed last year's more stringent version, but the current bill was shaped by a working group he convened, raising the stakes for his decision.
With the bill now on his desk, Newsom's signature would create the most significant AI safety law in the US and set a new precedent for holding developers accountable. A veto would reaffirm the industry's push for federal, rather than state-level, oversight.
Also on our radar: The push for AI governance extends beyond California, with New York's legislature recently passing its own AI safety bill. Meanwhile, SB 53 also includes provisions to establish CalCompute, a public cloud computing cluster aimed at fostering safer and more equitable AI research.
