With Anthropic’s endorsement, SB 53 gains momentum as the first serious state-level attempt to regulate frontier AI development — despite fierce industry and political opposition.
A Surprise Win for SB 53
In a major development for U.S. AI governance, Anthropic has formally endorsed California’s AI safety bill SB 53, a piece of legislation that would impose first-of-its-kind transparency rules on the world’s most powerful AI model developers.
- Anthropic’s endorsement is notable given the strong opposition SB 53 has faced from major tech lobbying groups like the Consumer Technology Association and Chamber for Progress.
- The bill, authored by California State Senator Scott Wiener, would apply to only the largest AI labs, including OpenAI, Google DeepMind, xAI, and Anthropic itself.
“The question isn’t whether we need AI governance — it’s whether we’ll develop it thoughtfully today or reactively tomorrow,” Anthropic wrote in a blog post.
What SB 53 Would Require
If enacted, SB 53 would introduce new legal requirements for companies building “frontier AI models” — systems powerful enough to pose catastrophic risks if misused.
Key provisions include:
- Mandatory safety and security reports before deployment
- Internal safety frameworks to prevent harmful misuse
- Whistleblower protections for employees reporting safety concerns
- Focus on mitigating extreme risks, such as:
- AI-assisted bioweapons development
- Cyberattacks
- Scenarios leading to mass casualties or $1B+ in damages
Notably, the bill does not focus on more common concerns like deepfakes or AI bias, instead targeting high-consequence risks from advanced models.
A Shift in Tone — and in Strategy
Unlike the more sweeping and controversial SB 1047, which Governor Newsom vetoed last year, SB 53 is narrower in scope, applying only to companies with over $500 million in revenue developing powerful foundation models.
- The bill has been amended to remove mandatory third-party audits, addressing a key concern from tech companies who argued audits were too burdensome.
- It was also shaped by advice from Governor Newsom’s AI advisory panel, which includes Stanford AI expert Fei-Fei Li.
“SB 53’s drafters have shown respect for technical reality and legislative restraint.” – Dean Ball, Foundation for American Innovation
Divisions Within the AI Industry
Anthropic’s support stands in sharp contrast to OpenAI’s lobbying. While OpenAI hasn’t publicly named SB 53, its chief global affairs officer wrote a letter to Governor Newsom opposing AI regulation that could “push startups out of California.”
- Critics like Miles Brundage, OpenAI’s former head of policy research, called the letter “misleading garbage”, arguing that SB 53 is both reasonable and necessary.
- Anthropic’s co-founder Jack Clark responded directly on X: “We have long said we would prefer a federal standard. But in the absence of that, this creates a solid blueprint for AI governance that cannot be ignored.”
Federal vs. State Debate Heats Up
Opponents of SB 53 — including Andreessen Horowitz, Y Combinator, and the Trump administration — argue that AI regulation should be left to federal authorities.
- Some claim SB 53 may violate the Commerce Clause, which prohibits state laws that interfere with interstate commerce.
- The Trump administration has even threatened to block state-level AI regulation altogether.
Still, Anthropic and other supporters argue that waiting for Washington is no longer viable.
“Powerful AI advancements won’t wait for consensus in Washington,” Anthropic stated.
The Road Ahead
SB 53 has already passed the California Senate, but still awaits a final Assembly vote before it can reach Governor Gavin Newsom’s desk.
- Newsom has not commented publicly on SB 53, though his previous veto of SB 1047 casts uncertainty over the outcome.
- Policy analysts believe the bill now stands a real chance of passing, especially with Anthropic’s support and recent revisions.
