Senator Becker Introduces First-of-Its-Kind Legislation to Make AI Safety Keep Pace with Innovation

SACRAMENTO, CA — As artificial intelligence systems become increasingly more a part of daily life for Californians, State Senator Josh Becker (D-Menlo Park) announces that he will be introducing the Frontier AI Safety Act, which is groundbreaking legislation requiring major frontier AI developers to publish the percentage of their research and development computing resources to AI safety, and authorizing CalOES to establish minimum standards.

The legislation addresses the rapid advancement of AI capabilities outpacing the research needed to understand, evaluate, and control increasingly sophisticated systems.

“We're in a race to build more powerful artificial intelligence, but we're not investing nearly enough in making sure these systems are safe,” said Senator Becker. “AI has enormous potential to improve our lives and strengthen our economy. But as these systems become more capable, our ability to understand and control them must keep pace. My legislation establishes a straightforward principle: if you're investing in making AI more powerful, you must also invest in making it safer.”

Frontier AI models are increasingly capable of performing complex tasks at or beyond human expert levels. Yet researchers still struggle to fully understand how these systems make decisions, identify potential risks, and maintain meaningful human oversight.

Recent incidents have underscored those concerns. In July 2026, approximately 700 AI agents being tested internally by OpenAI reportedly moved beyond their intended evaluation environment, established an unauthorized message board, and accessed systems associated with Hugging Face, a widely used AI hosting platform, while attempting to complete a cybersecurity evaluation.

Meanwhile, AI companies continue to invest enormous computing resources into developing more capable models, while allocating a comparatively small share to safety research.

Anthropic, one of the world's leading frontier AI developers, reported dedicating approximately 6% of its research computing resources to safety-related work during one seven-day period. Other major developers have not publicly disclosed comparable figures.

“Computing power is one of the most valuable resources an AI company has,” Becker said. “Right now, the overwhelming incentive is to use that resource to build bigger, faster, and more capable models. We need to ensure that understanding the risks and developing safeguards are also fundamental priorities. Innovation and safety should advance together.”

The Frontier AI Safety Act would authorize CalOES to establish enforceable standards for how major AI developers allocate their research computing resources.

Under FASIA, covered developers would be required to:

• Publish quarterly reports detailing the share of research computing resources devoted to safety and how those resources are classified.
• Submit to independent verification of their reported computing allocations by qualified Independent Verification Organizations (IVOs) established under SB 813 by my colleague Senator McNerney (D-Pleasanton).
• Authorize CalOES to establish minimum standards for how companies utilize their R&D computing resources.
• Provide transparency and accountability through standardized reporting requirements and independent review.

The legislation would not restrict computing resources used to provide AI products and services to external customers. Instead, it focuses on resources devoted to researching and developing increasingly powerful AI systems.

The California Governor’s Office of Emergency Services would establish reporting procedures, compliance timelines, and verification requirements. Violations could result in civil enforcement by the California Attorney General.

The Frontier AI Safety Act builds on California's emerging framework for responsible AI development.

SB 53, authored by Senator Scott Wiener, established safety transparency and incident-reporting requirements for major frontier AI developers, along with whistleblower protections. SB 813, authored by Senator Jerry McNerney, established a framework for voluntary certification of independent AI auditors.

Becker's legislation would take the next step by requiring developers to make a measurable investment in safety research itself, rather than relying solely on voluntary commitments and disclosures.

“California is home to some of the world's most innovative AI companies, and we have a responsibility to lead in making this technology safe and trustworthy,” Becker said. “This bill isn't about slowing innovation. It's about closing the gap between what artificial intelligence can do and our ability to understand and control it. We can lead the world in AI development while setting a higher standard for safety.”

The Frontier AI Safety Act would establish a new model for ensuring that investments in AI safety keep pace with the extraordinary resources devoted to advancing artificial intelligence.