Regulating AI at the State Level: California's Proposed Bill

The California AI Bill: What Is It?

Senior California lawmaker, state Senator Scott Wiener (D—San Francisco), acknowledges the pivotal role California, particularly Silicon Valley, plays in the AI ecosystem. Despite several initiatives at the federal level to draft AI legislation, Wiener believes that California, given its rich tech history, should not be a passive spectator.

While there’s a consensus about the need for federal AI regulatory protocols, Wiener insists on California’s proactive stance, especially considering the past shortcomings around technologies like social media.

What Does the Bill Propose?

At its core, the bill aims to regulate “frontier” AI systems that are on the cusp of significant capability enhancements. Key features include:

Transparency Requirements: AI systems that consume a yet-to-be-specified amount of computational power during their training phase would need to adhere to transparency norms.

Legal Liability: The bill underscores the accountability of entities failing to exercise due diligence, thereby resulting in unforeseen outcomes or enabling nefarious uses of AI.

Security Measures: To prevent advanced AI technologies from landing in unauthorized international territories, the bill advocates for rigorous security protocols.

CalCompute: Perhaps one of the most visionary components is the creation of “CalCompute,” a state research cloud envisioned to democratize access to AI infrastructure, especially for academia and startups.

Categorized as an intent bill, it currently provides an overarching view without delving into specifics. The hope is to catalyze conversations, refine the bill based on feedback, and pave the way for a comprehensive piece of legislation by 2025.

Unresolved Matters

A significant aspect yet to be ironed out relates to the determination of the computational threshold that would trigger the bill’s provisions. Moreover, the accountability conundrum – whether corporations or individual employees should be held liable – remains a point of contention.

Furthermore, Wiener admits the current lack of readiness within the California government to audit AI systems or enforce the bill.

Implications and Reception

The juxtaposition of being an AI hub and a regulatory authority is a delicate one for California. Governor Newsom’s recent executive order emphasizes a balanced approach to AI, ensuring that the state remains at the forefront of AI innovation while mitigating its risks.

This approach seems to resonate with global sentiments too. For instance, the U.K.’s stance on avoiding an overly restrictive AI regulatory framework is in line with California’s aspirations to foster innovation.

With states ramping up their efforts to outline AI boundaries and 25 states introducing AI bills as of July 2023, the race to AI governance is very much alive.

Sources:

Time

The Verge

Politico

‍