Big Tech opposes a new California bill, SB 1047, that mandates rigorous safety testing for advanced AI models costing over $100 million or requiring substantial computing power.
The legislation, introduced by State Senator Scott Wiener, also demands that developers incorporate a “kill switch” for AI systems and undergo third-party safety audits. Moreover, it empowers the state attorney general to take legal action against non-compliant developers, particularly when AI threatens critical government systems.
California legislators could vote on this comprehensive AI regulation bill as early as this week. Despite its passage in the state Senate and recent approval by the state Assembly appropriations committee, the bill has encountered significant opposition from Silicon Valley, including high-profile detractors like Nancy Pelosi and other Congressional Democrats from California. They argue that the bill could drive AI development out of the state and stifle innovation in open-source AI, which is crucial for advancing safer AI technologies more rapidly.
Tech leaders have expressed mixed reactions. While acknowledging the need for stricter AI regulations to prevent potential cyberattacks and other risks, companies like Google and Meta have raised concerns about the bill’s implications for AI development and research in California. OpenAI has suggested that federal regulation would be more appropriate, arguing that SB 1047 introduces legal uncertainties, particularly regarding open-source AI models.
Despite these concerns, the bill has supporters within the tech community, including AI pioneers like Geoffrey Hinton and Yoshua Bengio, who advocate for its potential to safeguard public interests against the rapid advancement of AI technology.