California has taken an important step towards regulating artificial intelligence with Gov. Gavin Newsom, where Gavin Newsom signs new state laws that require major AI companies. Many of them are headquartered in the state and have publicly disclosed plans to mitigate the potentially catastrophic risks posed by advanced AI models.
The law also begins the development of Calcompute, a government consortium tasked with creating mechanisms for reporting serious safety incidents, extending whistleblower protection to employees of AI companies, and creating public computing clusters for the research and innovation of safe, ethical, sustainable AI. By following these new rules, with compelling companies like Openai, Meta, Google Deepmind and Anthropic, California can effectively set standards for AI surveillance.
Newsom framed the law as a balance between encouraging public protection and innovation. In his statement, he wrote: “California has proven that it can establish regulations to protect our communities and ensure that the growing AI industry continues to thrive. This law balances that.”
Written by Senator Scott Wiener, the law follows a failed attempt to pass similar AI laws last year. Wiener said the new law known in Shorthand SB 53 (in the case of Senate Bill 53) focuses on transparency rather than deviations from the previous SB 1047 bill rejected by the newspapers last year.
“The SB 53 passages have achieved remarkable victories across California and the AI industry,” said Sunny Gandhi, Vice President of Political Affairs at Encode AI, co-sponsored by SB 53. The setting for a competitive, safe, and globally respected AI ecosystem. ”
Industry responses to new laws are split. Jack Clark, co-founder of AI company Anthropic, which supported SB 53, wrote to X: [the California Governor] For signature [Scott Wiener’s] SB 53 establishes transparency requirements for frontier AI companies. This helps to provide better data about these systems and the companies that build them. Humanity is proud to support this bill. “He emphasized that while federal standards remain important to prevent a patchwork of state regulations, California has created a framework that balances public safety with ongoing innovation.
Openai, who did not support the bill, added to the news outlet, “We are pleased that California has created an important path to harmony with the federal government, which is the most effective approach to AI security,” adding that if the law is implemented correctly, cooperation between the federal and state governments on the deployment of AI will be possible. Meta spokesman Christopher SGRO similarly told the media that he “supports balanced AI regulations,” calling SB 53 a “positive step in that direction,” and Meta looks forward to working with lawmakers to protect consumers while fostering innovation.
Despite being a state-level law, California law will have global reach as 32 of the world's top 50 AI companies are based in the state. The bill would require AI companies to report accidents to the California Department of Emergency Services, protect whistleblowers and allow engineers and other employees to raise safety concerns without putting their careers at risk. SB 53 also includes civil penalties for violations that can be enforced by the state attorney general, but AI policy experts like Miles Brundage have relatively weak penalties.
Brundage, formerly head of policy research at Openai, said in a post on X that SB 53 represents “A Step Forward,” but that the report required “actual transparency,” a stronger minimum risk threshold, and a technically robust third-party evaluation.
Collin McCune, head of government affairs at Andreessen Horowitz, also warned the law, “there is a risk of squeezing startups, slowing innovation and entrenching the biggest players,” and said it would set a dangerous precedent for state-by-state regulations that allow startups to create patchwork patchwork of 50 compliance administrations that have no navigation. Several AI companies that opposed the bill also made similar arguments.
California aims to promote transparency and accountability in the AI sector through its requirements for public disclosure and case reporting. However, critics like McCune have argued that the law could make compliance difficult to establish control of small businesses and Big Tech AI.
Thomas Woodside, co-founder of Secure AI Project, called startup concerns “exaggerated.”
“The bill only applies to companies that are training AI models on huge amounts of calculations that cost hundreds of millions of dollars. This is something small startups can't do,” he said. luck. “It's to report something very serious that's wrong, whistleblower protection, and a very basic level of transparency. No other obligations apply to businesses with annual revenues of less than $500 million.”
