california’s new ai safety law shows regulation and innovation don’t have to clash 

California’s AI Safety Law Demonstrates Regulation and Innovation Can Coexist

SB 53, the recently enacted California legislation concerning AI safety and transparency, serves as evidence that state-level regulation doesn’t necessarily impede advancements in artificial intelligence.

Insights from Encode AI’s Adam Billen

This assessment comes from Adam Billen, the vice president of public policy at Encode AI, a youth-focused advocacy organization. He shared his perspective on the latest episode of Equity.

Billen explained that policymakers recognize the necessity of addressing AI’s potential risks. He further stated they understand, through experience with numerous other issues, that legislation can be crafted to both safeguard innovation and ensure product safety.

Core Components of SB 53

SB 53 represents a pioneering effort as the nation’s first bill mandating transparency from major AI developers regarding their safety and security measures. Specifically, it focuses on preventing misuse, such as cyberattacks targeting vital infrastructure or the creation of bio-weapons.

The law also requires adherence to these established protocols, with enforcement handled by the Office of Emergency Services.

Industry Practices and the Need for Legislation

Billen pointed out that many companies are already implementing the practices outlined in the bill. This includes conducting safety testing and releasing model cards.

However, he acknowledged that some firms may be reducing their commitment to safety in certain areas. This underscores the importance of legislation like SB 53 to maintain standards.

Competitive Pressures and Safety Standards

Billen also highlighted a trend where some AI companies consider relaxing safety standards when facing competitive challenges. OpenAI, for instance, has indicated a willingness to “adjust” its safety requirements if a competitor introduces a high-risk system without comparable safeguards.

He argues that policy can reinforce companies’ existing safety commitments, preventing compromises driven by competition or financial considerations.

Opposition and the Broader Debate

Although public resistance to SB 53 was less pronounced than to its predecessor, SB 1047, which was vetoed last year, concerns within Silicon Valley and among AI labs remain. A common argument is that any AI regulation will ultimately hinder the United States in its competition with China.

This sentiment is reflected in substantial financial contributions from companies like Meta, venture capital firms like Andreessen Horowitz, and individuals like OpenAI president Greg Brockman to political action committees supporting pro-AI candidates.

Federal Preemption Efforts

Earlier this year, there were attempts to enact an AI moratorium, which would have prohibited states from regulating AI for a decade. Encode AI led a coalition of over 200 organizations to oppose this proposal, but Billen cautions that the battle is ongoing.

Senator Ted Cruz, a proponent of the moratorium, is pursuing alternative strategies to achieve federal preemption of state laws. He recently introduced the SANDBOX Act, allowing AI companies to seek waivers from federal regulations for up to ten years.

Concerns About Federal Standards

Billen also anticipates a forthcoming federal AI standard, presented as a compromise, but which he believes would effectively override state laws. He warned that such narrowly defined federal legislation could “delete federalism for the most important technology of our time.”

He emphasized that SB 53 is designed to address a specific set of concerns and shouldn’t be replaced by a sweeping federal mandate.

Focus on Specific Risks

Billen clarified that SB 53 isn’t intended to be a comprehensive solution to all AI-related challenges. He stated that it doesn’t aim to halt progress in the AI race with China.

He believes that state bills focusing on issues like deepfakes, transparency, algorithmic discrimination, children’s safety, and governmental AI use are valuable and shouldn’t be abandoned.

The Importance of Export Controls and Chip Production

Billen argued that if the goal is to surpass China in AI development, efforts should focus on measures like export controls and bolstering domestic chip production. He noted that these are the types of initiatives the industry should be advocating for.

Legislative proposals such as the Chip Security Act, aiming to prevent the flow of advanced AI chips to China, and the CHIPS and Science Act, designed to increase domestic chip manufacturing, are examples of such efforts.

Industry Reluctance and Conflicting Signals

However, some major tech companies, including OpenAI and Nvidia, have expressed reservations or opposition to certain aspects of these initiatives, citing concerns about effectiveness, competitiveness, and security.

Nvidia, in particular, has a strong financial interest in maintaining sales to China, a historically significant market. Billen suggested that OpenAI might be hesitant to fully support chip export controls to avoid jeopardizing its relationship with key suppliers like Nvidia.

The Trump administration’s inconsistent messaging further complicates the situation, with a temporary reversal of an export ban on advanced AI chips to China in exchange for a percentage of revenue.

A Democratic Process in Action

Billen concluded that SB 53 exemplifies a successful democratic process, where industry and policymakers collaborate to reach a mutually acceptable agreement. He described it as “very ugly and messy,” but ultimately a testament to the foundations of the country’s economic system.

He believes SB 53 provides a strong example of how regulation and innovation can coexist.

This article was first published on October 1.