September 30, 2024 – California Governor Gavin Newsom has vetoed the controversial “Safe Innovation for Frontier Artificial Intelligence Model Act” (SB 1047), citing multiple factors that influenced his decision. These include the burden the bill would place on AI companies, California’s leading position in the field, and criticisms that the bill may be overly broad.
The SB 1047 bill was designed to establish safety standards for developers of large-scale artificial intelligence models, specifically those with a training cost exceeding $100 million or reaching a certain level of computing power. It required developers of such models to take precautionary measures such as pre-deployment testing, simulated hacking attacks, installing cybersecurity protections, and providing whistleblower protections.
The bill, which was introduced to the California legislature in February, sparked significant debate. On August 19th, the legislature released a revised version of the bill. While it received support from prominent figures in the AI community such as Geoffrey Hinton and Yoshua Bengio, it also faced opposition from scientists including Yann LeCun and Fei-Fei Li, as well as leading AI companies, who argued that it would “hinder innovation.”
In his veto message, Newsom stated, “While SB 1047 starts with a noble intention, it fails to consider whether AI systems are deployed in high-risk environments, involve critical decision-making, or use sensitive data. Instead, it applies strict standards even to the most basic functions of large-scale system deployments. I do not believe this is the best way to protect the public from the real threats posed by this technology.”
Newsom further argued that the bill could “give the public a false sense of security about controlling this rapidly evolving technology.” He added, “Smaller, more specialized models may be as dangerous or even more so than the models targeted by SB 1047 – and the cost could be the weakening of innovation that drives progress in the public interest.”
The governor expressed his agreement that safety protocols and guardrails should be established, and that bad actors should face “clear and enforceable” consequences. However, he emphasized that the state should not “settle for a solution that is not based on an empirical trajectory analysis of AI systems and capabilities.”
Jamie Radice, Meta’s Public Policy Manager, responded, “We are pleased that Governor Newsom has vetoed SB 1047. This bill would have stifled AI innovation, harmed business growth and job creation, and broken the state’s long tradition of promoting open-source development. We support responsible AI regulation and remain committed to working with legislators to promote better approaches.”