California's senate has officially passed a significant AI safety proposal that establishes new disclosure rules for major tech firms.
According to the bill’s sponsor, state senator Scott Wiener, SB 53 “mandates that leading AI labs reveal their safety measures, introduces protections for whistleblowers within AI companies, and establishes a public cloud resource to broaden computing access (CalCompute).”
The legislation is now headed to Governor Gavin Newsom of California, who will decide whether to sign it into law or reject it. While Newsom has yet to make a public statement on SB 53, he previously vetoed a broader safety measure from Wiener last year but did approve narrower laws targeting issues such as deepfakes.
At that time, Newsom recognized the necessity of “shielding the public from serious risks posed by this technology,” but argued that Wiener’s earlier bill was overly restrictive, imposing “strict requirements” on large AI models even when they were not “used in high-risk contexts, making critical choices, or handling sensitive information.”
Wiener noted that feedback from an expert AI policy advisory group, which Newsom assembled after the previous veto, helped shape the current bill.
Politico adds that SB 53 has recently been revised to require developers of advanced “frontier” AI models with annual revenues under $500 million to share only general safety information, while those with higher revenues must submit more comprehensive reports.
A range of Silicon Valley tech companies, venture capitalists, and advocacy organizations have spoken out against the bill. In a recent communication to Newsom, OpenAI did not refer to SB 53 directly but stated that to avoid “overlap and conflict,” businesses should be considered compliant with California's safety laws if they already follow federal or EU regulations.
Andreessen Horowitz’s AI policy chief and top legal executive also recently warned that “many of the current state AI proposals — such as those in California and New York — could exceed constitutional boundaries in terms of state regulation of interstate commerce.”
The firm’s co-founders had earlier cited tech policy as a reason for supporting Donald Trump’s re-election campaign. Following this, Trump’s administration and his supporters advocated for a decade-long moratorium on state-level AI regulations.
Meanwhile, the company Anthropic has publicly endorsed SB 53.
“We have always expressed a preference for nationwide standards,” Anthropic’s co-founder Jack Clark commented in a recent post, “but without that, this bill offers a robust model for governing AI that deserves attention.”
