Senate Bill 53, known as the “Transparency in Frontier Artificial Intelligence Act,” has recently become law in California after being signed by Governor Gavin Newsom. The bill was authored by Senator Scott Wiener and comes after a previous version, SB 1047, was vetoed last year due to concerns that it might hinder innovation within the AI sector. The earlier bill aimed to mandate extensive risk testing for AI creators, particularly those with models costing over $100 million to develop.
Following the veto, Governor Newsom sought input from AI researchers to create a more balanced approach, resulting in a report that informed the development of SB 53. The new legislation incorporates several recommendations including requirements for large AI companies to disclose their safety protocols, the implementation of whistleblower protections, and increased transparency with the public.
Under SB 53, large AI developers must publish a framework detailing how they adhere to various safety and security standards on their websites. Any updates to safety protocols must also be made public within 30 days. However, the bill’s effectiveness in safeguarding whistleblowers is subject to interpretation, as many companies favor voluntary guidelines that lack enforceable penalties.
Additionally, the law establishes mechanisms for reporting critical safety incidents to California’s Office of Emergency Services and outlines civil penalties for noncompliance, with enforcement by the Attorney General’s office. The California Department of Technology is tasked with recommending annual updates to the law based on stakeholder input and technological advancements.
The reception among AI companies has been mixed, with many initially opposing the bill, citing concerns about potential negative effects on the industry’s presence in California. The state’s significant population and tech hubs provide it with substantial influence over AI regulations.
Amid these debates, some companies, like Anthropic, have supported the bill, while others, such as Meta and OpenAI, have actively lobbied for alternative frameworks. OpenAI has proposed that compliance might be achieved through adherence to federal or international regulations, suggesting a potential pathway for AI developers to navigate state requirements.
Source: https://www.theverge.com/ai-artificial-intelligence/787918/sb-53-the-landmark-ai-transparency-bill-is-now-law-in-california