California is advancing a new AI safety bill, placing the state at the forefront of AI regulation amidst a deadlocked Congress. The bill, known as the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, has ignited a debate between the state's tech leaders and lawmakers.
Proponents of the bill argue that AI's rapid development necessitates immediate safeguards to protect the public. Conversely, opponents fear the legislation will hinder AI innovation.
California State Senator Scott Wiener, the bill's author, highlighted the resistance from the tech sector. “In the tech sector, there are people who just don’t want any regulation,” Wiener said, emphasizing the necessity of proactive measures.
The bill mandates that developers of advanced AI models conduct rigorous safety testing and certify their products. It grants the state’s attorney general the authority to take legal action if AI technologies cause significant harm, particularly any mass casualty events. This legislation targets AI models with substantial computational and financial resources behind them. A key provision requires developers to have a "kill switch" to shut down potentially harmful AI systems. Additionally, it establishes a Frontier Model Division within the state's Government Operations Agency to oversee and enforce these standards.
Over 140 AI startup founders, represented by Y Combinator, expressed concerns that the bill could drive AI talent away from California. Critics also worry about its impact on the development of open-source AI models.
Senator Wiener responded to these concerns, clarifying that the bill sets thresholds for advanced models that are not yet in existence. He addressed misinformation about the bill’s implications, stating it does not target AI startups or penalize developers with jail time. Wiener acknowledged the importance of the open-source tech community and assured that their concerns are being considered.
The debate touches on fundamental questions about AI’s potential dangers, the necessity of regulation, and whether regulation stifles innovation. Some prominent AI experts, including Geoffrey Hinton and Yoshua Bengio, have supported the bill, citing the existential risks AI could pose.
Wiener pointed out that California is stepping up due to inaction at the federal level, though he hopes Congress will eventually address AI regulation. “I think in many areas it’s ideal to have federal standards, but I’m not confident that Congress will act on AI regulation,” Wiener stated, referencing Congress’s inaction on federal data privacy law and social media regulation.
The bill has passed the state assembly’s Privacy Committee and is set for consideration by the Judiciary Committee next week. It is part of a broader legislative effort in California to address various AI-related issues, including privacy, data transparency, copyright, and bias.