California Senate Bill 1047, titled the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, was a proposed state law aimed at regulating the development and deployment of large-scale artificial intelligence systems. Introduced by State Senator Scott Wiener in February 2024, the bill sought to impose safety requirements on developers of AI models above certain computational thresholds. After passing the California State Legislature in August 2024, the bill was vetoed by Governor Gavin Newsom on September 29, 2024, marking a significant moment in the ongoing debate over AI regulation in the United States.
The veto was a culmination of intense lobbying and public debate, drawing responses from major technology companies, AI researchers, and advocacy groups. Proponents argued that the bill was necessary to prevent potential catastrophic harms from advanced AI, while opponents, including many in the tech industry, contended that it would stifle innovation and impose unrealistic obligations. The aftermath of the veto has led to renewed discussions about alternative regulatory approaches at both state and federal levels.
Legislative Background
Senator Scott Wiener introduced SB 1047 in February 2024, with the goal of establishing a regulatory framework for "frontier AI models" - those trained with computing power exceeding 10^26 floating-point operations (FLOPs). The bill required developers to implement safety protocols, including the ability to promptly shut down models, and to report serious incidents to a newly created state agency. It also mandated that developers conduct pre-deployment testing and ensure that their models could not be easily modified to cause mass casualties or enable cyberattacks.
The bill underwent several amendments during its legislative journey. Notably, an amendment in August 2024 narrowed the scope by removing provisions that would have created a new state agency, instead assigning oversight to the Department of Technology. Another amendment exempted models developed primarily for academic or research purposes, provided they were not used for commercial deployment.
Legislative Passage
SB 1047 passed the California State Assembly on August 28, 2024, by a vote of 41-16, and the State Senate on August 30, 2024, by a vote of 30-8. The bill received bipartisan support, with some Republicans joining Democrats in favor. The passage was seen as a landmark moment, as California is home to many of the world's leading AI companies, including OpenAI, Anthropic, and Google DeepMind.
Governor's Veto
On September 29, 2024, Governor Gavin Newsom vetoed SB 1047, issuing a statement that expressed support for AI innovation but criticized the bill's approach. Newsom argued that the bill's focus on model size and computing power was too narrow and could give a false sense of security, while potentially deterring the development of beneficial AI applications. He emphasized the need for a more flexible, evidence-based regulatory framework that could adapt to the rapidly evolving technology.
The veto was not entirely unexpected, as Newsom had earlier signaled concerns about overregulation. In his veto message, he noted that the bill did not account for the diverse applications of AI and could impose burdensome requirements on smaller developers. He also pointed to ongoing federal efforts and California's own executive order on AI, signed in September 2023, as more appropriate venues for addressing these issues.
Reactions to the Veto
Reactions to the veto were sharply divided. Supporters of the bill, including Senator Wiener and various AI safety advocacy groups, expressed disappointment. Wiener called the veto a setback for public safety and vowed to continue pushing for regulation. Some AI researchers, such as Yoshua Bengio (though not directly linked, he is a prominent figure), had publicly supported the bill, arguing that precautionary measures were essential.
Opponents, including many in the tech industry, praised the veto. Companies like OpenAI and Anthropic had lobbied against the bill, arguing that it would hinder innovation and that the thresholds were arbitrary. Venture capital firms and startup advocates also welcomed the decision, claiming that the bill would have driven AI companies out of California.
Aftermath and Regulatory Landscape
The veto has not ended the debate over AI regulation in California. Senator Wiener has indicated plans to introduce revised legislation in the next session, incorporating feedback from stakeholders. Meanwhile, Governor Newsom has convened an expert panel to study AI risks and make recommendations for future policy. The state has also continued to implement its executive order, which focuses on responsible AI procurement and risk assessment.
At the federal level, the veto has intensified calls for a national AI regulatory framework. The Biden administration had issued an executive order on AI in October 2023, but comprehensive legislation remains pending in Congress. The lack of federal action has led to a patchwork of state initiatives, with California's approach being closely watched as a potential model.
Industry and Academic Perspectives
The veto has sparked discussions within the AI community about the best ways to ensure safety without stifling progress. Some researchers argue that self-regulation and voluntary commitments, such as those made by leading labs to pause development of the most powerful models, are insufficient. Others contend that regulation should focus on specific high-risk applications rather than on the models themselves.
Anthropic and OpenAI have both established internal safety teams and have called for government oversight, but they opposed SB 1047's specific provisions. Google DeepMind has also been active in AI safety research, but did not take a public stance on the bill. Academic institutions like Stanford AI Lab and Berkeley AI Research have contributed to the debate through research and policy recommendations.
Future Prospects
As of early 2025, no new state AI bill has been passed in California, but discussions continue. The debate over SB 1047 has highlighted the challenges of regulating a technology that evolves rapidly and has broad applications. The outcome of future legislative efforts will depend on balancing innovation with public safety, and on the ability to craft rules that are both effective and flexible.
In the absence of binding regulation, many AI developers have adopted voluntary safety frameworks, such as the "frontier safety framework" proposed by Anthropic. These efforts, while not legally enforceable, represent a step toward responsible development. The veto of SB 1047 has thus become a case study in the complex interplay between technology, policy, and public opinion.