California lawmakers pass AI safety bill SB 53 — but Newsom could still veto

The Staatssenate of California gave final approval Early on Saturday morning to a large AI safety account that sets new transparency requirements for large companies.

When described by the author, Senator Scott Wiener, State state, SB 53 “Requires that large AI labs are transparent about their safety protocols, creates whistleblower protection for [employees] At AI Labs & creates a public cloud to expand compute access (calcompute). “

The bill now goes to Californian governor Gavin Newsom to sign or veto. He did not respond publicly to SB 53, but last year he pronounced Veto for a more extensive safety bill that was also written by Wiener, while signing narrower legislation that focuses on issues such as Deepfakes.

At the time, Newsom recognized the importance of “protecting the public against real threats of this technology”, but criticized Wiener’s earlier account for applying “strict standards” to large models, regardless of whether they were “deployed in risky environments, [involved] Critical decision -making or the use of sensitive data. “

Wiener said that the new account was influenced by recommendations from a AI expert policy panel that Newsom gathered after his veto.

Polico also reports That SB 53 was recently changed, so that companies that develop “Frontier” AI models, while producing less than $ 500 million in annual income, only have to reveal safety details at a high level, while companies that make more have to offer more detailed reports.

The bill has been opposed by a number of Silicon Valley companies, VC companies and lobby groups. In A recent letter to NewsomOpenAi did not mention SB 53 specifically, but argued that ‘duplication and inconsistencies’ should be considered compatible with safety rules throughout the state, as long as they meet the federal or European standards.

And Andreessen Horowitz’s head of AI policy and Chief Legal Officer recently claimed That “many of today’s state of AI accounts – such as proposals in California and New York – risk a line” by violating constitutional boundaries for the way states can regulate trade between states.

The co-founders of A16Z had previously pointed to the technical regulation as one of the factors that led them to withdraw Donald Trump’s offer for a second term. The Trump administration and its allies then asked for a ban of 10 years on the AI ​​regulation of the state.

Anthropic has since been released in favor of SB 53.

“We have said for a long time that we would prefer a federal standard,” said anthropic co-founder Jack Clark in a post. “But in the absence, this creates a solid blueprint for AI governance that cannot be ignored.”