Governor Newsom's Veto of SB 1047: A Flawed Approach to AI Regulation

California Governor Gavin Newsom’s veto of Senate Bill 1047 (SB 1047) underscores a growing tension between rapid AI advancements and rushed legislative efforts to regulate the technology. The bill, which would have mandated strict safety protocols for AI models developed with over $100 million in investment, was touted as a crucial step in protecting the public from potential harms of large-scale AI systems. However, Newsom's decision to reject it highlights deeper issues with the legislative framework.

The bill's provisions, including mandatory safety tests and built-in "kill switches" for human intervention, were designed to address the potential risks of advanced AI models. Yet Newsom criticised its broad approach, claiming that the focus on expensive, large-scale models was too simplistic. In his veto letter, he noted that the bill created a "false sense of security" by targeting only major AI systems, while smaller, specialised models could still pose significant threats, especially in sensitive environments. As he pointed out, SB 1047 did not account for whether AI systems were deployed in high-risk environments or involved critical decision-making.

This veto comes amidst a wider legislative push in California, where 17 bills have been introduced to address AI-related risks such as disinformation, deepfakes, political manipulation, and racial bias in algorithms. Newsom acknowledged these broader efforts, stating, 

“This year, the legislature sent me several thoughtful proposals to regulate AI.” 

His recognition of these issues highlights the ongoing debate about the role AI plays in undermining democratic processes and spreading dangerous misinformation.

However, SB 1047, according to Newsom, missed the mark in addressing the complexity of AI regulation. The bill’s reliance on the cost of developing AI models as a regulatory threshold raised significant concerns. By focusing on expensive models, lawmakers risked ignoring the dangers posed by smaller systems that could operate under the radar, but in equally high-stakes environments. Newsom noted that the bill "applies stringent standards to even the most basic functions so long as a large system deploys it,” a broad-brush approach that failed to address the nuanced risks AI can pose.

Critics, particularly from the tech industry, welcomed Newsom's veto. Jennifer Everett, a partner in Alston & Bird’s technology and privacy group, remarked that 

"frontier model developers that would have been most directly impacted by the law are probably breathing a sigh of relief." 

Meanwhile, opponents of the veto, including AI safety advocates, view it as a setback in establishing crucial safeguards to prevent AI from causing catastrophic harm.

State Senator Scott Wiener, the bill's author, was quick to respond, accusing tech giants like Y Combinator and Andreessen Horowitz of running a "propaganda campaign" to mislead the public about the bill's intentions. Wiener stated,

“I’ve had tough bills before, and bills where there have been misinformation. I’ve never had a bill with this level of misinformation.”

In reality, Newsom’s veto reflects a broader dilemma: how to balance the need for AI regulation with the risk of stifling innovation. While SB 1047 is now off the table, California’s efforts to regulate AI are far from over. The veto hints at a more nuanced regulatory approach, one that focuses on the specific risks posed by AI deployments rather than blanket measures based on model size or development costs. With 32 of the world’s leading AI companies based in California, the state’s decisions will have profound implications for the future of AI governance—not just in the U.S., but globally.