"Silicon Valley Meets Sacramento: California's AI Regulation Tango"

On September 10th, legal professionals and economic leaders gathered at Orrick’s law offices in San Francisco for a seminar titled "Navigating the AI Legislative Landscape: SB 1047 and Beyond."

The event explored the complex intersection of AI innovation and the urgent need for effective regulatory frameworks in California. The panel, consisting of Orrick senior partners, executives from the Bay Area Council and Silicon Valley Leadership Group kicked off with one of the most controversial and largely contested bills to come out of Sacramento in recent days - California AI Senate Bill 1047, titled the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act (“SB 1047” or the “bill”).

SB 1047: A Balancing Act

California's proposed SB 1047, amended several times by the state’s senate due to strong resistance from some industry groups, aims to mitigate risks associated with high-impact AI models such as, 1) models trained using a quantity of computing power greater than 10 ^ 26 integer - the cost to train these frontier models can run over $100,000,000 based on market prices, and 2) AI models created by fine tuning a covered model with cost of fine tuning in excess of $10,000,00.  

Some of the requirements of the bill for developers of covered models include:

• Implement safety protocols to prevent unauthorized use, screen models for cybersecurity and infrastructure risks,

• Implement a full shut down in the “face of a critical harm,”  

• Assess whether a model could cause or enable a critical harm, and implement appropriate safeguards,

• AI labs will be required to submit public statements outlining their safety testing practices,

• Developers must retain a third-party auditor to assess developers compliance annually 

The bill's prescriptive nature, however, has sparked debate within the tech industry with Google, Meta, and OpenAI arguing that the bill will stifle AI innovation and damage open models development by regulating the process of model development. At the same time, Anthropic and Elon Musk’s xAI have announced measured support for SB 1047 arguing that the benefits will likely outweigh the costs.

The "Sacramento Effect": California's Proactive Stance

The panel also noted that California is emerging as a trailblazer in AI legislation, building upon the foundation of the CCPA (California Consumer Privacy Act). The "Sacramento Effect" underscores the state's commitment to proactively address the challenges and opportunities presented by AI advancements. Further amplifying the "Sacramento Effect", the Governor's 2023 Executive Order establishes a strategy for state agencies to both recognize the risks associated with AI and accelerate its adoption within the government. It emphasizes individual agency responsibility in risk mitigation while encouraging the use of AI to improve state services. This proactive approach demonstrates California's dedication to harnessing AI's potential while managing its challenges.

Other Broader AI Legislative Bill

Bills such as SB 942 and SB 2013 seek to increase transparency and accountability by mandating labeling of AI-generated content and disclosure of training data sources. Key provisions of the bills include:

• Mandates providers of large generative AI systems to label AI-generated content (images, videos, audio) with both visible and embedded disclosures.

• Requires covered providers to create an AI detection tool, allowing users to query whether content was created by AI.

• Requires developers of generative AI systems made available to the public in California to disclose information about the data used to train these systems.

The Bay Area: A Thriving AI Ecosystem

• Investment & Job Creation: The Bay Area remains a global hub for AI investment and job creation, with a diverse ecosystem spanning across infrastructure, foundation models, and industry-specific applications. Big Tech is pouring massive amounts of investments in AI startups. Thus far Microsoft alone has invested over $13 billion and owns 49% stake in OpenAI the last is now valued at $150 billion. Google has invested about $2 billion in Anthropic. The Google-Anthropic collaboration involves incorporating Claude 3, which is Anthropic’s frontline technology, into Google’s products. Amazon has invested about $4 billion in Anthropic and has become the main cloud provider for the AI company. Venture capital funds such as Sequoia Capital, Founders Fund, Andreessen Horowitz, Tiger Global and others have also poured billions into Silicon Valley’s AI rapidly emerging ecosystem. 

• Workforce Readiness: California and NVIDIA have recently teamed up to advance AI education and innovation. Their joint initiative aims to provide accessible AI learning opportunities for Californians, foster job growth in the AI sector, and leverage AI to address pressing state challenges. This collaboration will bring NVIDIA's AI resources into community colleges, including curriculum, certifications, hardware, software, and workshops, enabling students, educators, and workers to acquire new skills and thrive in the AI-powered future. Policymakers are actively looking to address concerns about potential job displacement and economic disparities arising from AI adoption.

The Road Ahead for SB 1047 and AI Regulation

The Orrick event provided valuable insights into California's evolving AI regulatory landscape as well as AI focus areas. As the state continues to lead in shaping AI policy, striking the right balance between innovation, safety, and ethical considerations will be paramount. The ongoing conversation and collaboration between policymakers, industry leaders, and the public are key to ensuring a responsible and beneficial AI future.

These initiatives showcase California's holistic approach to AI development, addressing not only technological advancement but also environmental sustainability, ethical considerations, and workforce preparation. As these programs unfold, they are set to reinforce California's status as a pioneer in the AI landscape.

What’s Next?

It’s important to note that the future of these initiatives is not yet set in stone. Governor Newsom must either sign, approve without signing, or veto all four measures by the end of September. Should any be enacted into law, they will add to the growing number of state laws imposing new affirmative duties on the development and use of AI models, systems, and applications. Moreover, California's actions may inspire other states to adopt similar measures, potentially leading to a wave of AI-focused legislation across the country.

The panel voted 4 to 1 against California Bill 1047, stating it's unlikely Governor Newsom will sign it in its current form. They cited concerns about the bill's potential negative effects on California's economy, which could harm largely Democratic legislators in the polls. Additionally, there's an opposition from influential groups in the tech industry, policy circles, and academia.

Catch you next time!