Gov. Newsom vetoes pioneering AI safety legislation in California

Gov. Newsom Vetoes Pioneering AI Safety Legislation in California

SACRAMENTO, Calif. — California Governor Gavin Newsom has vetoed a historic bill aimed at instituting the nation’s first safety measures for large-scale artificial intelligence (AI) models. This decision, made public on Sunday, represents a significant setback for advocates pushing for regulatory oversight in a rapidly advancing industry that currently operates with minimal constraints.

The proposed legislation, known as Senate Bill 1047, was intended to introduce some of the earliest regulations on AI systems in the United States. Supporters of the bill, including tech figures like Elon Musk and organisations such as Anthropic, saw it as a foundational step towards transparency and accountability in an industry where even developers are often unclear on the intricate behaviours of their models.

Gov. Newsom, however, cited concerns that the bill’s broad requirements would stifle innovation and unduly burden the industry. “While well-intentioned, SB 1047 does not take into account whether an AI system is deployed in high-risk environments, involves critical decision-making or the use of sensitive data,” Newsom stated. He argued that imposing stringent standards on all large AI systems, regardless of their use case, was not the optimal strategy for addressing public safety threats posed by AI technology.

The vetoed legislation would have required companies to test their AI models rigorously and disclose their safety measures publicly to prevent potential misuse, such as disruptions to the electrical grid or assistance in developing chemical weapons. The bill also included provisions for whistleblower protections for employees within these companies.

State Senator Scott Wiener, the bill’s author, expressed disappointment over Newsom’s decision. Describing the veto as a “setback,” Wiener underlined the growing risks associated with AI developments. “The companies developing advanced AI systems acknowledge that the risks these models present to the public are real and rapidly increasing,” Wiener said. Although Wiener vowed to continue advocating for AI safety, he noted the progress made in elevating the issue through the legislative debate.

In response to the veto, Gov. Newsom announced plans to collaborate with industry experts to create guidelines for powerful AI models. This advisory council will include prominent figures like Fei-Fei Li, an AI pioneer who opposed SB 1047, suggesting a more flexible and adaptive approach to AI safety.

The legislative effort in California comes amid broader attempts to regulate AI technology at both the state and federal levels. California lawmakers passed several bills this year addressing AI issues, including deepfakes and worker protections. These efforts were grounded in lessons learned from previous failures to regulate social media corporations effectively.

The debate in California also echoes larger global dynamics, particularly as the United States lags behind Europe in implementing comprehensive AI regulations. The European Union has already taken significant steps to mitigate AI risks, setting a benchmark that American legislation has yet to meet.

The bill targeting large AI models specifically focused on systems requiring more than $100 million to develop. While no current AI models meet this threshold, experts argued that such investments could become common as the industry scales up rapidly. Daniel Kokotajlo, a former researcher at OpenAI, remarked on the vast power concentrated within private companies and the associated risks.

Advocates of the bill maintained that it would have brought much-needed oversight to an industry fraught with concerns over job displacement, misinformation, privacy violations, and biased automated systems. Critics, including former U.S. House Speaker Nancy Pelosi, voiced fears that the bill would deter innovation and harm California’s thriving tech industry. Many AI developers and the California Chamber of Commerce had lobbied against the bill, arguing that it would inhibit the development and sharing of open-source software.

Gov. Newsom has positioned California as a leader in AI adoption, with plans to integrate generative AI tools across various state functions, including transportation, tax services, and homelessness programmes. The state announced a collaboration with Nvidia to train students, faculty, developers, and data scientists in AI technologies. Recent legislative achievements include stringent laws to combat election-related deepfakes and measures protecting Hollywood workers from unauthorised AI usage.

Despite the veto, the debate in California has inspired other state lawmakers to consider similar regulatory measures. Tatiana Rice, deputy director of the Future of Privacy Forum, indicated that efforts to legislate AI safety are likely to persist across the country.

This development underscores the evolving challenges that governments face as they seek to balance the benefits of technological innovation with the imperative to safeguard public interests.

Source: Noah Wire Services