A group of esteemed academics has thrown their weight behind California's groundbreaking AI safety legislation as it approaches a critical juncture in the state's lawmaking process. In an exclusive letter obtained by TIME, four distinguished professors — Yoshua Bengio, Geoffrey Hinton, Lawrence Lessig, and Stuart Russell — have voiced their strong support for the proposed bill.

The legislation in question, known as the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, was introduced earlier this year by Senator Scott Wiener. It aims to establish rigorous safety protocols for companies developing large-scale AI models, mandating thorough testing for potential hazards and the implementation of comprehensive risk mitigation strategies.

In their letter, the experts emphasize the pressing need for such regulation, pointing out the stark contrast between the current oversight of AI systems and other less potentially impactful industries. They argue that the bill represents the absolute minimum necessary for effective governance of this rapidly advancing technology.

The professors' appeal is directed to key figures in California's legislative process, including the leaders of the state senate and assembly, as well as Governor Gavin Newsom. The bill has already cleared the senate and now faces a crucial vote in the assembly later this month. If successful, it will then move to the governor's desk for final approval or veto by the end of September.

This endorsement from leading figures in the field underscores the growing consensus on the need for proactive measures to ensure the safe development of AI technologies. As California continues to lead in technological innovation, this bill could set a significant precedent for AI regulation nationwide and beyond.

California's proposed AI safety legislation has emerged as a focal point in the broader debate over AI regulation, given the state's economic clout and its position as a hub for leading AI developers. With federal action stalled and potential policy reversals looming, California's role in shaping AI governance has taken on heightened significance.

The bill, if enacted, would have far-reaching implications for companies operating within the state. It has garnered substantial public support, according to polls, but faces strong opposition from certain industry groups and tech investors. Critics argue that the legislation could impede innovation, negatively impact the open-source community, and potentially cede AI leadership to other countries.

Prominent voices in the tech world have aligned against the bill. Venture capital firm Andreessen Horowitz has launched a campaign urging citizens to oppose the legislation. Other notable figures, including YCombinator representatives, Meta's Chief AI Scientist Yann LeCun, and Stanford professor Fei-Fei Li, have also expressed concerns.

The controversy primarily centers on provisions requiring developers to provide assurances against "critical harms" from their AI models, such as facilitating the creation of weapons of mass destruction or causing severe damage to critical infrastructure. However, the bill's scope is limited to systems that meet specific cost and computing power thresholds, likely affecting only the largest AI developers.

Lennart Heim, a researcher at the RAND Corporation, has noted that no currently existing system would fall under the bill's purview, highlighting the forward-looking nature of the legislation.

This debate underscores the complex balancing act between fostering innovation and ensuring public safety as AI technology continues to advance rapidly.

A group of esteemed experts in artificial intelligence and technology law have lent their voices to support California's proposed AI safety bill. Their collective expertise and reputation add significant weight to the ongoing debate about AI regulation.

The letter's authors, who include Turing Award winners Yoshua Bengio and Geoffrey Hinton, often referred to as "godfathers of AI," assert that the risks associated with advanced AI systems are both probable and significant enough to warrant safety testing and precautionary measures.

Stuart Russell, author of a widely recognized standard textbook on AI, and Lawrence Lessig, a Harvard Law professor known for his pioneering work in Internet law and the free culture movement, round out this distinguished group of supporters.

Their concerns extend beyond the previously mentioned risks to include the potential dangers posed by autonomous AI agents operating without human oversight. This highlights the broad spectrum of potential issues that advanced AI systems could present.

Yoshua Bengio, in a statement to TIME, expressed worry that market competition and profit-seeking might prevent technology companies from adequately addressing these risks on their own. He emphasized the need for regulatory frameworks to guide the development of frontier AI technologies.

The involvement of these respected figures underscores the growing recognition within the academic and research communities of the need for proactive measures to ensure the safe and responsible development of AI technologies.

The letter in support of California's AI safety bill addresses several key criticisms and misconceptions about the proposed legislation. It emphasizes that the bill's scope is limited to only the largest AI models, countering concerns about stifling innovation across the board.

The authors point out that many large AI developers have already voluntarily committed to implementing safety measures similar to those outlined in the bill. They also note that comparable regulations in Europe and China are actually more stringent, positioning this bill as a balanced approach.

A significant aspect of the bill highlighted in the letter is its provision for robust whistleblower protections. This is seen as increasingly crucial given reports of potentially reckless behavior in some AI labs.

Senator Wiener, the bill's sponsor, has noted recent amendments made in response to feedback from the open-source community. These changes include exemptions for original developers from shutdown requirements once a model is out of their control, and limitations on liability for significantly modified versions of their models.

Despite these adjustments, some critics maintain that the bill would require open-source models to have a "kill switch," a claim that appears to be contested.

The letter characterizes the bill as a "light-touch piece of legislation" relative to the potential risks, noting that it doesn't include a licensing regime or require government permission for model training. Instead, it relies on self-assessments of risk.

The authors conclude with a strong statement, suggesting that failing to implement these basic measures would be a "historic mistake," underscoring their view of the bill's importance in the evolving landscape of AI governance.

Lawrence Lessig, a prominent legal scholar and one of the letter's authors, has emphasized the unique position California holds in potentially shaping the future of AI regulation. In an email statement, Lessig highlighted the opportunity that Governor Gavin Newsom has to establish California as a pioneer in AI governance.

Lessig underscored the urgency of legislative action in this domain, pointing out that California's role is particularly crucial given its status as home to a significant number of leading AI firms. This concentration of industry leaders within the state makes it an ideal testing ground for regulatory frameworks that could influence AI development on a broader scale.

The professor's comments reflect a belief that California's actions could have far-reaching implications beyond its borders. By taking the initiative to implement thoughtful regulation, the state could set a precedent for other regions grappling with similar challenges posed by rapidly advancing AI technologies.

Lessig's statement aligns with the letter's overall message, reinforcing the idea that proactive measures are necessary to ensure the responsible development of AI. It also suggests that California's potential leadership in this area could help address the current regulatory gap at the national level.

This perspective from a respected figure in technology law adds another dimension to the ongoing debate about the proposed AI safety bill, emphasizing its potential to position California at the forefront of AI regulation.

Here's a conclusion for the text:

The support from renowned experts for California's AI safety bill underscores the growing recognition of the need for proactive regulation in the rapidly evolving field of artificial intelligence. Their endorsement brings significant credibility to the proposed legislation and highlights the potential risks associated with unchecked AI development.

The bill, if passed, could position California as a pioneer in AI governance, potentially influencing similar efforts across the nation and beyond. This legislation represents a delicate balance between fostering innovation and ensuring public safety, addressing concerns from various stakeholders including the tech industry, open-source community, and public interest groups.

As the bill moves through the final stages of California's legislative process, it faces both strong support and opposition. The outcome of this legislative effort could have far-reaching implications for the future of AI development and regulation, not just in California but globally.

Ultimately, the debate surrounding this bill reflects the broader challenges society faces in harnessing the potential of AI while mitigating its risks. As AI continues to advance, the decisions made now in California may well shape the trajectory of this transformative technology for years to come.