New AI Bill Introduced in California by SB 1047 Author

California State Senator Scott Wiener, who stirred the pot with last year's contentious AI safety bill SB 1047, is back with another bill that's set to ruffle feathers in Silicon Valley. On Friday, Wiener introduced SB 53, which aims to shield employees at major AI labs from retaliation if they raise concerns about their company's AI systems posing a "critical risk" to society. Additionally, the bill proposes the creation of CalCompute, a public cloud computing cluster designed to provide researchers and startups with the computing power needed to develop AI for the public good. Wiener's previous bill, SB 1047, sparked intense national debate over how to manage large AI systems that could potentially lead to catastrophic events like significant loss of life or cyberattacks causing damages over $500 million. Despite the controversy, Governor Gavin Newsom vetoed the bill in September, arguing it wasn't the right approach. The fallout from SB 1047 was fierce. Silicon Valley bigwigs argued that the bill would undermine America's edge in the global AI race, fueled by what they saw as unfounded fears of AI-induced apocalyptic scenarios. Wiener hit back, accusing some venture capitalists of orchestrating a "propaganda campaign" against his bill. Notably, Y Combinator's claim that SB 1047 could jail startup founders was labeled misleading by experts. SB 53 seems to be a strategic pivot, focusing on the less contentious elements of SB 1047. It emphasizes whistleblower protections and the establishment of CalCompute. Wiener isn't backing away from existential AI risks, though. SB 53 clearly protects whistleblowers who believe their employers are developing AI systems that could pose a "critical risk," defined as a foreseeable or material risk that could result in the death or serious injury of over 100 people, or more than $1 billion in damages. The bill targets frontier AI model developers like OpenAI, Anthropic, and xAI, prohibiting them from retaliating against employees who report concerns to California's Attorney General, federal authorities, or other employees. These developers must also report back to whistleblowers on internal processes flagged as concerning. As for CalCompute, SB 53 would set up a group including University of California representatives and other researchers to plan its development. They'll decide the size of the cluster and who gets access to it. SB 53 is still in the early stages of the legislative process, needing approval from California's legislative bodies before reaching Governor Newsom's desk. Silicon Valley's response to this new bill will be closely watched. Passing AI safety bills in 2025 might be trickier than in 2024, when California passed 18 AI-related bills. The AI safety movement seems to be losing steam, especially after Vice President J.D. Vance's comments at the Paris AI Action Summit, where he emphasized AI innovation over safety. While CalCompute could be seen as promoting AI progress, the future of legislative efforts focusing on existential AI risks remains uncertain. [ttpp][yyxx]

Related article

Comments (29)