California Governor Gavin Newsom has vetoed the controversial “Frontiers in Artificial Intelligence Model Safety Innovation Act,” also known as SB 1047. In his veto statement on September 30, Newsom cited several factors influencing his decision, including the burden the bill would impose on AI companies, California’s leading role in the industry, and concerns over the legislation’s potentially broad scope.
SB 1047 was designed to set safety standards for developers whose training costs exceed $100 million or meet certain computational power thresholds, aiming to ensure the secure development of large-scale AI models. The legislation required developers of cutting-edge AI models to implement preventive measures, such as conducting pre-deployment testing, simulating cyberattacks, installing cybersecurity safeguards, and providing protection for whistleblowers.
The bill was introduced to the California legislature in February and sparked significant controversy. An amended version was released by the state legislature on August 19. The bill received backing from prominent figures in the AI community, including Geoffrey Hinton and Yoshua Bengio, often referred to as the “godfathers of AI.” However, it faced opposition from other scientists, including Fei-Fei Li and Kai-Fu Lee, as well as from leading AI companies, who argued the legislation would stifle innovation.
In his statement, Newsom acknowledged that while the intentions behind SB 1047 were good, it failed to consider whether AI systems were deployed in high-risk environments or involved critical decision-making processes with sensitive data. He noted that the bill would impose strict standards even on the most basic functionalities of large systems, stating, “I do not believe this is the best way to protect the public from the true threats posed by this technology.”
He further expressed concerns that the legislation could give the public a false sense of security regarding the control of rapidly evolving AI technologies. Newsom commented, “Smaller, more specialized models might pose equal to or even greater dangers than those targeted by SB 1047—the cost of which could weaken innovation that drives public interest progress.”
While Newsom stated his agreement on the necessity of establishing safety protocols and ensuring clear consequences for bad actors, he argued against settling for a solution that does not rely on empirical analysis of AI systems and their capabilities.
Meta’s public affairs manager, Jamie Radice, welcomed Newsom’s veto, stating, “We are pleased that Governor Newsom vetoed SB 1047. The bill would have stifled AI innovation, hampered business growth and job creation, and undermined the state’s long-standing tradition of promoting open-source development. We support responsible AI regulations and remain committed to working with legislators to foster better approaches.”
