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

California State Senate gave the final approval Early Saturday morning, to a major AI security bill, setting new transparency requirements on large companies.

As described by its author, The state senator Scott Wiener, SB 53 “Requires that large AI laboratories be transparent on their safety protocols, creates denunciation protections for [employees] At Ai Labs and creates a public cloud to extend access to calculations (Calcompute). »»

The bill now goes to the Governor of California Gavin Newsom to sign or oppose his veto. He did not publicly comment on SB 53, but last year, he vetoed a larger security bill also written by Wiener, while signing closer legislation targeting issues like Deepfakes.

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

Wiener said the new bill had been influenced by the recommendations of an IA expert policy panel that Newsom summoned after his veto.

Politico also reports The fact that SB 53 has been recently changed so that companies developing “border” AI models while providing less than $ 500 million in annual income will only have to disclose high -level security details, while companies by manufacturing more than that will have to provide more detailed reports.

The bill opposed a number of companies in Silicon Valley, venture capital companies and lobbying groups. In A recent letter to NewsomOPENAI did not mention the SB 53 in particular, but argued that to avoid “duplication and inconsistencies”, companies should be considered as compliance with state -scale safety rules as long as they meet federal or European standards.

And the head of the AI ​​policy of Andreessen Horowitz and the legal director recently claimed that “many bills on today’s state AI – such as proposals in California and New York – risk” crossing a line by violating the constitutional limits on the way states can regulate interstate trade.

The co-founders of A16Z had previously highlighted technological regulations as one of the factors leading them to support Donald Trump’s offer for a second term. The Trump administration and its allies then called for a 10 -year ban on state AI regulations.

Anthropic, meanwhile, came out in favor of SB 53.

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