California has long been at the center of AI innovation. Home to 35 of the global top 50 AI companies and San Francisco and San Jose alone providing one quarter of the world’s AI patents and research, California, being a leader in innovation and oversight, comes as no surprise. Yet, as AI continues to develop at lightning speed, lawmakers are trying to figure out how to keep it safe, equitable, and accessible without slowing innovation down.
The Emergence of AI—and Concern among the Public
AI is now an integral part of daily life, driving everything from recommendations based on individual behavior to virtual assistants. But public awareness hasn’t kept up. While approximately 77% of digital devices today depend on some type of AI, very few in the population can describe how it functions. As the impact of AI has increased, so has public unease: between 2021 and 2023, the percentage of Americans concerned about AI rose from 37% to 52%, and just 10% indicated they were more thrilled than apprehensive about its uses.
This growing worry put additional pressure on policymakers, particularly in California, where the stakes are highest, to set guardrails without strangling innovation.
SB-1047: An Ambitious Try at AI Regulation
One of the most ambitious attempts by the state was in the guise of SB-1047, the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, proposed by Senator Scott Wiener. The legislation aimed to regulate only the most powerful and resource-intensive AI systems—those costing more than $100 million to train.
Its major proposals were:
- Cybersecurity standards must be mandated to avoid abuse.
- Risk assessments before release and annual third-party audits.
- “Kill switch” functionality for AI models that could threaten critical systems.
- Whistleblower protections for AI employees are raising safety concerns.
- CalCompute: a state-run cloud infrastructure to democratize access to high-performance computing.
It was welcomed as a progressive measure by supporters. Economic Security California led the support for CalCompute for breaking up Big Tech’s stranglehold and making the AI field accessible to startups and researchers. More than 100 workers and graduates from large AI corporations even called on Governor Newsom to support the bill’s whistleblower protections.
Resistance from Tech Giants and Scholars
Even with the support, SB-1047 was met with strong resistance. Industry giants such as Google, Meta, and OpenAI threatened that the bill would drive innovation out of California. Top scholars, including Stanford’s Fei-Fei Li, contested the bill as stifling open-source AI creation and unreasonably burdening startups.
Others went so far as to claim the bill did not go far enough, citing ambiguous language and loopholes that could potentially allow large businesses to opt out.
Finally, Governor Newsom vetoed the bill, citing implementation issues and the requirement for additional study. However, he left open the possibility of future regulation, creating a working group of experts to develop new proposals. Their proposals are due shortly.
SB 53: A Targeted Follow-Up
Senator Wiener came back to the table with SB 53, a targeted bill constructed around two of the most popular concepts out of SB-1047: CalCompute and whistleblower protections.
SB 53 will:
- Shield workers who report potential safety issues in AI from retaliation.
- Launch CalCompute to provide cheap computing access for research universities, startups, and researchers.
Wiener put it most succinctly: “The greatest innovations occur when our most brilliant minds have the resources they require and the liberty to tell them how it is.” Encode’s VP of Political Affairs, Sunny Gandhi, concurred, labeling SB 53 a “blueprint for democratizing access while ensuring safety isn’t traded for speed.”
Why These Issues Matter
Whistleblowers have become an essential component of AI safety, providing uncommon insight into the inner workings of large AI companies. As innovators such as Geoffrey Hinton and Yoshua Bengio have cautioned, unregulated AI development poses significant threats, and insiders could be the first to notice when things start to go awry.
Meanwhile, training cutting-edge AI models takes massive amounts of computer processing resources, usually beyond the reach of small players. By establishing CalCompute, California aims to even the playing field and preclude innovation from being concentrated in the hands of tech titans.
California’s Influence on AI Policy
As California leads the way, others around the nation and globe are taking notice. New York just spent $90 million on developing its own AI innovation center, hoping to catch up with California’s leadership. Nationally, other countries like Brazil, Chile, and Canada are closely observing California’s policy trials, searching for a model to implement at home.
At the same time, Washington is slower. It has made incremental moves—moving against AI-driven robocalls and writing agency regulations—but still does not have a single national AI law, forcing states such as California to lead the way.
What’s Next?
The path from SB-1047 to SB 53 shows how hard it is to control an evolving and dynamic technology. California continues to seek the proper balance: promoting innovation while providing public safety and accountability.
As the months go on, more reports and policies come out, the world—and California—will be paying close attention. If the state can balance things just right, it may serve as a model for 21st-century responsible, inclusive AI governance.
