By Jodie Godoy
(Reuters) – California Gov. Gavin Newsom signed state law on Monday to show requirements that CHATGPT developer Openai and other major players will unveil plans to mitigate potential catastrophic risks from cutting-edge AI models.
California is home to top AI companies such as Openai, Alphabet’s Google, Meta Platforms, Nvidia and Anthropic, and with the law it seeks to lead regulations for industries that are important to the economy.
“California is proving that we can establish regulations to protect our communities and ensure that the growing AI industry continues to thrive,” Newsom said in a press release.
Newsom’s Office said the law, known as SB 53, fills the gap left by the US Congress, which has not passed extensive AI laws so far, and provides a model that the US should follow.
Newsom said if federal standards are introduced, the state legislature must “ensure consistency with all standards while maintaining the hiber established by SB 53.”
Last year, Newsom rejected its first attempt at California’s AI law, which was facing a fierce industry pushback. The bill spent more than $100 million on AI models, requiring that third-party auditors be hired annually to confirm risk assessments, allowing the state to impose fines in hundreds of millions of dollars.
The new law requires companies with revenues of over $500 million to assess the risk that cutting-edge technology can be freed from human control, assisted in the development of biological weapons, and allowed them to be made public. Each violation will be fined up to $1 million.
Jack Clark, co-founder of AI company humanity, is called “a strong framework that balances public safety with continuous innovation.”
The industry still wants a federal framework to replace California law and others recently enacted in Colorado and New York. Last year, some Republicans in the U.S. Congress voted in Senate 99-1 for some bids to help states prevent AI regulation.
“The biggest risk of SB 53 is setting precedents for states taking the lead in managing the national AI market, not the federal government. Startups will create a patchwork of 50 compliance regimes that do not have the resources to navigate.”
The representative US Republican in California, Jay Obernolté, is working on AI laws that can preempt several state laws, and his office declined to comment further on the pending law.
Some Democrats are also debating how to establish federal standards.
“Will you regulate AI, do 17 states want to do that, or do they want to do that in Congress?” US representative Ted Liu, a Democrat from Los Angeles, said at a recent hearing on AI law in the U.S. House of Representatives.
(Reporting by Jodie Godoy of New York, edited by Chris Sanders and Edmund Kraman)
