In the early hours of Saturday morning, the California State Senate officially granted its final approval to Senate Bill 53 (SB 53), a sweeping piece of legislation that establishes new transparency obligations for major technology companies developing advanced artificial intelligence systems. This measure, widely recognized as one of the most ambitious state-level attempts to regulate the rapidly advancing AI sector, seeks to strike a balance between fostering innovation and addressing the pressing societal risks that this technology may pose if left unchecked.
The legislation was introduced and championed by State Senator Scott Wiener, who carefully crafted the bill to respond to concerns raised by both policymakers and the public about the potential misuse or unsafe deployment of powerful AI models. According to Wiener, the bill requires that large-scale AI research laboratories disclose information about the safety mechanisms and precautionary measures they have put in place. It further establishes essential protections for employees who act as whistleblowers should they uncover internal malpractices or hidden dangers within these organizations. Additionally, the bill proposes the creation of a publicly accessible cloud computing infrastructure, known as CalCompute, which aims to broaden access to high-performance computing resources for researchers and innovators who might otherwise be excluded due to limited financial or technical means.
Now that the California legislature has approved it, the bill proceeds to the desk of Governor Gavin Newsom, who will ultimately decide whether to sign it into law or exercise his authority to veto it. While Newsom has not yet issued public comments specifically addressing SB 53, his past decisions provide important context. Last year, he vetoed an earlier AI safety proposal from Senator Wiener, which was considerably broader in scope and imposed uniformly strict standards on all large AI models, regardless of the risk environment in which they were deployed. However, during the same legislative session, Newsom did approve more narrowly tailored regulations, such as those targeting harmful applications of AI in generating deceptive media, including deepfakes. At that time, he publicly acknowledged the necessity of safeguarding citizens from the genuine risks presented by emerging AI developments, but he voiced concerns that Wiener’s initial proposal extended overly stringent regulatory constraints even on systems that did not interact with sensitive information or operate in high-stakes contexts.
Senator Wiener has stated that the revised bill reflects crucial input from a panel of prominent AI experts that Governor Newsom himself convened after last year’s veto. This advisory group offered recommendations intended to ensure that California adopts a regulatory framework that is both rigorous in terms of safety standards and pragmatic with regard to fostering economic competitiveness. Recent amendments to SB 53, reported by Politico, illustrate this balancing effort. Specifically, smaller companies developing so-called frontier AI models—those at the cutting edge of capability—but generating less than $500 million in annual revenue will only be required to provide broad and high-level descriptions of their safety processes. In contrast, companies earning revenues above this threshold will face more rigorous obligations to furnish detailed and comprehensive safety reports. In this way, the legislation attempts to scale requirements in proportion to both the size and potential societal impact of the organizations involved.
Despite its safeguards, the bill has encountered strong opposition from influential actors within the state’s powerful technology ecosystem. A coalition of Silicon Valley companies, well-known venture capital firms, and industry lobbying organizations has argued that the proposal threatens to create overlapping, inconsistent, and burdensome regulatory obligations. For instance, in a recent letter addressed to Governor Newsom, OpenAI refrained from naming SB 53 directly but contended that to prevent costly duplication and regulatory contradictions, firms should be deemed compliant with California’s rules as long as they already adhere to similar national or international frameworks, such as those established by U.S. federal agencies or European Union authorities.
Additional criticism has come from Andreessen Horowitz, one of Silicon Valley’s most prominent venture capital firms. The company’s head of AI policy and its chief legal officer argued that proposed laws of this nature—whether drafted in California, New York, or elsewhere—risk crossing constitutional boundaries by attempting to regulate interstate commerce, a power traditionally reserved for the federal government. This concern reflects the continuing tension between state-level attempts to fill regulatory vacuums and the broader constitutional design of national governance.
The debate over state-level regulation has also intersected with national politics. The co-founders of Andreessen Horowitz previously identified what they characterized as overbearing state regulation of technology firms as a significant factor behind their decision to lend support to Donald Trump’s effort to secure a second presidential term. In alignment with these views, the Trump administration and its political allies have called for an outright halt to state-based AI regulation, advocating instead for a federal moratorium lasting a decade.
Nevertheless, SB 53 has not been universally criticized within the technology industry. In fact, it has garnered active support from Anthropic, a leading AI research company that has positioned itself as a vocal advocate for responsible innovation. Jack Clark, one of the company’s co-founders, explained in a public statement that while Anthropic would prefer to see Congress and federal agencies implement a single national framework, the absence of such action makes California’s legislation a valuable interim model. Clark described SB 53 as offering a pragmatic blueprint for AI governance, one that demonstrates both urgency and foresight in the face of technological acceleration, and he emphasized that its provisions will likely influence future debates at the national level.
In sum, California’s SB 53 represents a pivotal moment in the ongoing effort to regulate artificial intelligence: an ambitious yet carefully revised piece of legislation that ventures beyond traditional government caution and boldly addresses both economic and safety concerns. While its ultimate fate rests in the hands of Governor Newsom, the bill has already sparked vigorous debate among policymakers, industry leaders, and the broader public. It has simultaneously underscored the urgent need for oversight in this transformative technological field and highlighted the deep disagreements over how such governance should be structured, whether at the state, national, or global level.
Sourse: https://techcrunch.com/2025/09/13/california-lawmakers-pass-ai-safety-bill-sb-53-but-newsom-could-still-veto/
