California’s latest AI safety fight has reached the governor’s desk. State lawmakers gave final approval early on Saturday morning to SB 53, a bill that would set new transparency expectations for large companies working on advanced AI systems.
The next move belongs to California governor Gavin Newsom. He has not commented publicly on SB 53, and the bill can still become law or be vetoed.
What SB 53 Is Designed To Do
SB 53 is centered on disclosure and accountability. As described by its author, state senator Scott Wiener, the bill requires large AI labs to be transparent about safety protocols, creates whistleblower protections for employees at AI labs and establishes a public cloud project known as CalCompute.
That makes the bill different from a simple product rule. It is not described as banning a particular AI tool or requiring one narrow technical fix. Instead, it focuses on whether major AI developers should publicly explain how they manage safety and whether employees should have protection when they raise concerns.
CalCompute adds another piece to the proposal. According to Wiener’s description, it would create a public cloud intended to expand access to compute. In plain terms, that points to a debate not only about AI risks, but also about who can access the computing resources needed to build and study powerful AI systems.
The bill’s three core ideas can be read together:
- AI safety transparency: large AI labs would need to disclose information about their safety protocols.
- Whistleblower protections: employees at AI labs would receive protections when raising concerns.
- CalCompute: a public cloud would be created to expand compute access.
Why Newsom’s Decision Matters
SB 53 now goes to Gavin Newsom to sign or veto. That step is especially important because Newsom rejected a more expansive AI safety bill last year, also authored by Wiener.
At the same time, Newsom signed narrower legislation that targeted issues like deepfakes. That history suggests the decision may turn on how he views the scope of SB 53: broad enough to address real AI risks, but not so broad that it repeats the concerns he raised about Wiener’s earlier proposal.
Last year, Newsom acknowledged the importance of protecting the public from real threats posed by the technology. But he criticized the previous bill for applying stringent standards to large models regardless of whether those models were deployed in high-risk environments, involved critical decision-making or used sensitive data.
Wiener has said the new bill was influenced by recommendations from a policy panel of AI experts that Newsom convened after that veto. That detail matters because it frames SB 53 as a revised approach rather than a simple reintroduction of the earlier effort.
The Amendment That Changed Disclosure Burdens
Politico reports that SB 53 was recently amended to create different disclosure expectations based on company revenue. Companies developing "frontier" AI models while bringing in less than $500 million in annual revenue would only need to disclose high-level safety details.
Companies making more than that would need to provide more detailed reports. The result is a bill that still targets transparency, but does not appear to treat every developer of frontier AI models in exactly the same way.
That distinction is central to the political and business debate around AI regulation. Supporters of state-level rules want meaningful disclosure from companies building powerful systems. Critics often argue that too many overlapping obligations could create confusion, duplication or burdens that vary from state to state.
How The AI Industry Is Split
SB 53 has drawn criticism from a number of Silicon Valley companies, VC firms and lobbying groups. The objections are part of a broader pushback against state efforts to regulate AI.
OpenAI recently sent a letter to Newsom that did not mention SB 53 specifically. In that letter, OpenAI argued that to avoid duplication and inconsistencies, companies should be considered compliant with statewide safety rules if they meet federal or European standards.
Andreessen Horowitz has also criticized state AI bills. The firm’s head of AI policy and chief legal officer recently claimed that many state AI bills, including proposals in California and New York, risk crossing a line by violating constitutional limits on how states can regulate interstate commerce.
The political backdrop is broader than SB 53 alone. Andreessen Horowitz’s co-founders had previously pointed to tech regulation as one factor behind their decision to back Donald Trump’s bid for a second term. The Trump administration and its allies later called for a 10-year ban on state AI regulation.
Anthropic, by contrast, has come out in favor of SB 53. Anthropic co-founder Jack Clark said the company would prefer a federal standard, but argued that, in its absence, the bill creates a solid blueprint for AI governance that cannot be ignored.
The Immediate Stakes
The central question is now whether California will move ahead with its own AI safety framework or wait for broader federal rules. SB 53 sits directly inside that tension.
For supporters, the bill offers a way to require more openness from large AI labs, protect employees who speak up and broaden access to compute through CalCompute. For critics, it is part of a state-by-state regulatory push that could create duplication, inconsistency or constitutional concerns.
Newsom’s choice will decide the bill’s near-term future. If he signs it, California would move forward with new AI safety transparency rules. If he vetoes it, the debate will return to the same unresolved question: how much AI governance should come from states when federal standards are still the preferred path for some major AI companies.