A controversial AI safety bill known as SB 1047 gained approval from the California legislature and now awaits a decision from Governor Gavin Newsom. He has until September 30 to either sign or veto the legislation, which seeks to enforce strict safety protocols for newly developed AI models, while empowering the state attorney general to pursue legal actions against developers who fail to comply.
Industry Opposition and Support
The proposal has drawn criticism from tech giants like Google, Microsoft-backed OpenAI, and Meta Platforms. These firms caution that the law could hamper innovation and discourage AI development within the state. Meanwhile, Tesla CEO Elon Musk, who also leads the AI firm xAI, has expressed support for the initiative, emphasizing the necessity of safety protocols in the AI sector.
SB 1047 calls for rigorous safety assessments of advanced AI systems, especially those with development costs exceeding $100 million or which require considerable computing resources. The bill also mandates the inclusion of a “kill switch” to deactivate malfunctioning AI models. Developers must engage third-party auditors to scrutinize their safety practices and implement protections for whistleblowers.
Legislative Process and Political Dynamics
The State Assembly’s endorsement of the bill is a significant stride toward the regulation of AI technology in California. A procedural vote on recent amendments will be conducted by the State Senate this week. Governor Newsom’s decision will be influenced by strong lobbying from tech leaders, investors, academics, and federal legislators.
Should SB 1047 become law, it could set a national standard for AI regulation akin to the European Union’s AI Act. The bill’s sponsor, Democratic State Senator Scott Wiener, contends that proactive measures are required to mitigate potential risks posed by AI. The legislation’s aim is to address threats like AI systems compromising essential services such as power grids.
Tech Pushback
The Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, as SB 1047 is officially known, sets forth that developers of “frontier” AI models—those costing at least $100 million to train—must implement safety measures and testing frameworks. Companies must undergo audits and provide “reasonable assurance” that their models will not cause catastrophic events.
Top AI organizations and tech firms, including OpenAI, Anthropic, Cohere, and Meta, argue that these rigorous standards might hinder innovation and drive talent away from California. Y Combinator, a venture capital accelerator known for mentoring startups, has joined the debate. In a letter signed by 140 startups, Y Combinator criticizes the bill’s demands, fearing they could hamper the growth of new companies.
Anthropic is one company that has raised concerns about SB 1047. In a letter to Axios, the company voiced its stance that, while ensuring safe AI development is a worthy objective, the bill has significant shortcomings. OpenAI’s Chief Strategy Officer, Jason Kwon, cautioned in a letter to Senator Wiener that this bill could hamper growth, stall innovation, and cause engineers and entrepreneurs to relocate.
Some of the feedback has been taken on board with recent amendments to the bill. The AI Safety Act has undergone significant changes. The attorney general’s powers have been reduced, the Frontier Model Division has been scrapped, and AI developer responsibilities have been lessened. The bill now focuses on preventing future AI disasters and regulating AI models through the Board of Frontier Models.
Last Updated on November 7, 2024 3:05 pm CET
