California’s SB 53 has reached a decisive point. The state senate recently gave final approval to the AI safety bill, sending it to Governor Gavin Newsom, who can either sign it or veto it.
The measure matters because it is aimed at the biggest AI companies, not the broader startup ecosystem. It also arrives after Newsom vetoed another AI safety bill last year, SB 1047, which was also written by state senator Scott Wiener.
A narrower bill after last year’s veto
SB 53 is being judged partly against the shadow of SB 1047. Wiener’s earlier bill drew significant pushback, and Newsom ultimately vetoed it. The new bill is narrower, with a clear focus on large AI companies making more than $500 million in annual revenue.
That narrower design is central to why Max Zeff, speaking on TechCrunch’s flagship podcast Equity, believes the bill has a better shot at becoming law. The measure has also been endorsed by AI company Anthropic, which gives it a different political profile from the previous effort.
The core change is scope. Rather than putting broad requirements across the AI field, SB 53 tries to direct its heaviest obligations toward companies with major revenue from AI models. In the discussion, OpenAI and Google DeepMind were named as examples of the type of large companies the bill is trying to reach.
What SB 53 would require
The bill is presented as a practical check on the power of major AI labs. According to the TechCrunch discussion, it would require those labs to publish safety reports for their models. That would create a more formal public record of how the companies describe and evaluate their own systems.
SB 53 would also require companies to report incidents to the government. The source does not define every kind of incident that could qualify, but the basic direction is clear: the bill would move certain AI safety problems out of purely internal company processes and into a government reporting channel.
Another major piece concerns employees. If workers at covered AI labs have concerns, SB 53 would give them a way to report those concerns to the government without facing company pushback, even if they have signed NDAs.
Taken together, the bill’s central mechanisms are straightforward:
- Safety reports for models from covered AI developers.
- Government reporting when an incident occurs.
- A protected channel for employee concerns inside major AI labs.
- A focus on companies making more than $500 million in annual revenue from AI models.
Those elements explain why the bill is being framed as more than a symbolic gesture. It does not attempt to settle every question around AI regulation, but it would create concrete duties for the largest players in the field.
Why California is the key state
The state-level setting is not incidental. Kirsten Korosec argued on Equity that California matters because it is a hub of AI activity. As she put it, every major AI company is pretty much, if not based in California, operating with a major footprint there.
That gives California legislation a broader significance than a typical state bill might have. If the state places rules on major AI developers, those rules can touch companies that sit at the center of the industry’s current growth.
The argument is not that other states are irrelevant. The point is that California’s position in the AI economy makes its legislative choices especially important. A bill aimed at AI labs in California is likely to intersect with many of the companies building and deploying powerful AI models.
This is also why the bill’s focus on large companies is politically important. One criticism of the earlier legislative effort was that it could harm the startup ecosystem, which was described as a booming part of California’s economy. SB 53 attempts to answer that concern by applying its main requirements to larger AI developers instead of run-of-the-mill startups.
The startup carve-out and its limits
SB 53 does not appear to ignore smaller startups entirely. Anthony Ha noted in the discussion that smaller startups would still have to share some safety information, but not nearly as much as the larger companies covered by the bill’s main requirements.
That distinction is one of the bill’s defining trade-offs. It is designed to put meaningful obligations on major AI labs while reducing the burden on smaller companies. In practice, that makes the $500 million threshold one of the most important facts about the proposal.
The narrower design may also make the bill more complicated. Kirsten raised the issue of exceptions and carve-outs, asking whether the new bill is more complicated than the previous bill even though it is narrower. Max answered that in some ways, yes, because the measure tries specifically not to apply to small startups.
That complexity reflects the political challenge behind AI safety legislation. Lawmakers are trying to address companies that may become some of the most powerful in the world, while avoiding rules that could be seen as slowing smaller firms before they reach scale.
A state bill in a federal fight
SB 53 is also moving through a changed national environment. Anthony pointed to the broader landscape around AI regulation and said one major difference between last year and this year is the new president.
According to the discussion, the federal administration has taken much more of a no-regulation stance, favoring the idea that companies should be able to do what they want. The administration has also included language in funding bills saying states cannot have their own AI regulation.
Anthony said he does not think any of that has passed so far, but that it could potentially be attempted again in the future. That makes SB 53 part of a larger conflict over who gets to set the rules for AI: states such as California, or the federal government.
For now, the immediate question is narrower and concrete. California’s state senate has approved SB 53, and Governor Gavin Newsom must decide whether to sign or veto it. If signed, the bill could become one of the more meaningful state-level checks yet placed on the largest AI companies.