Kalifornijas likumdevēji pieņem AI drošības likumprojektu SB 53 – bet Newsom joprojām varētu veto

The California State Senate gave final approval early Saturday morning to a major AI safety bill, imposing new transparency requirements on large companies.

As described by its author, State Senator Scott Wiener, SB 53 “Requires large AI labs to be transparent about their security protocols, creates whistleblower protections for [employees] at AI Labs & creates a public cloud to expand Compute Access (Calcompute).”

The bill now goes to California Governor Gavin Newsom to sign or veto. He has not publicly commented on SB 53, but last year he also vetoed a broader security bill authored by Wiener while signing a narrower piece of legislation aimed at issues like DeepFakes.

At the time, Newsom acknowledged the importance of “protecting the public from the real threats posed by this technology,” but criticized Wiener’s previous bill for applying “strict standards” to large models regardless of whether they were “deployed in high-risk environments [involving] critical decision-making or the use of sensitive data.”

Wiener said the new bill is influenced by a policy group of AI experts that Newsom convened after his veto.

Politico also reports that SB 53 was recently amended so that companies developing “Frontier” AI models while generating less than $500 million in annual revenue will only be required to disclose high-level security information, while companies that generate more than that will be required to provide more detailed reports.

The bill has been opposed by several Silicon Valley companies, VC firms and lobbying groups. In a recent letter to Newsom, Openai did not specifically mention SB 53, but argued that to avoid “duplication and inconsistencies,” companies should be considered compliant with statewide security regulations as long as they meet federal or European standards.

And Andreessen Horowitz’s AI policy head and chief legal officer recently argued that “many of today’s state AI bills — such as proposals in California and New York — risk” crossing the line by violating constitutional boundaries on how states can regulate interstate commerce.

The A16Z co-founders had previously cited technology regulation as one of the factors that led them to support Donald Trump’s bid for a second term. The Trump administration and its allies later called for a 10-year ban on government AI regulation.

Meanwhile, anthropic has come out in favor of SB 53.

“We’ve said for a long time that we would prefer a federal standard,” Anthropic co-founder Jack Clark said in a post on The Local. “But if it doesn’t, it creates a solid AI governance plan that can’t be ignored.”