California Governor Gavin Newsom vetoed Senate Bill 1047 on Sunday, halting what would have been the United States' most sweeping regulatory framework for large-scale frontier artificial intelligence models. Authored by Democratic State Senator Scott Wiener of San Francisco, the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act had ignited months of intense lobbying across Silicon Valley, Washington, and academia.
Under SB 1047, developers training foundation models that require more than $100 million in computing power and exceed 10 to the 26th power integer or floating-point operations would have been required to conduct pre-deployment catastrophic risk testing, implement a full emergency 'kill switch' protocol, and face statutory liability from the California Attorney General if their systems caused mass casualties or over $500 million in critical infrastructure damage.
Why Newsom Rejected Compute-Threshold AI Regulation
In his formal veto message to the California State Legislature, Newsom emphasized that while California refuses to abandon proactive oversight of emerging technologies, SB 1047 relied on a rigid mathematical threshold rather than evaluating the actual context in which an AI model operates. He cautioned that smaller, specialized models deployed in sensitive sectors could pose greater real-world perils than massive general-purpose models performing benign tasks.
The legislation exposed a deep philosophical divide within the artificial intelligence industry. OpenAI, Meta, Google, venture capital firm Andreessen Horowitz, and former House Speaker Nancy Pelosi opposed the bill, warning it would stifle open-source weights, chill academic innovation, and drive AI startups out of California. Conversely, Anthropic, xAI founder Elon Musk, Turing Award laureates Geoffrey Hinton and Yoshua Bengio, and Hollywood labor unions backed the measure.
“While well-intentioned, SB 1047 applies stringent standards to even the most basic functions so long as a large system deploys it, rather than focusing on actual high-risk uses and empirical threat analysis.”
Silicon Valley Split and New Empirical Guardrail Initiative
Alongside the veto, Newsom announced that he has enlisted Stanford University computer scientist Dr. Fei-Fei Li, former U.S. Chief Data Scientist DJ Patil, and UC Berkeley Provost Jennifer Chayes to lead a new state initiative creating empirical, workable guardrails for frontier model evaluation prior to the next legislative session.
Newsom also directed California's Office of Emergency Services and the Department of Technology to expand risk assessments for catastrophic AI threats to the state's energy grid, water supply, and cybersecurity infrastructure. Earlier this month, the governor signed more than a dozen targeted AI bills addressing election deepfakes, digital performer replicas, and healthcare automation.
Comments (0)
Log in to join the discussion.