California lawmakers have recently amended a bill that seeks to impose new restrictions on artificial intelligence (AI), potentially setting a precedent for how tech companies develop and deploy AI systems. The bill, known as SB 1047, has been a focal point of debate among tech giants, policymakers, and academics, as it could introduce some of the first state-level safety regulations for AI technologies.
On Thursday, the State Assembly’s Appropriations Committee endorsed the amended version of SB 1047. The bill, authored by Senator Scott Wiener, mandates that companies must test the safety of powerful AI technologies before releasing them to the public. If these technologies result in significant harm—such as mass property damage or human casualties—the California attorney general would be empowered to take legal action against the companies responsible.
Industry Response and Amendments
The tech industry has been sharply divided over the bill. Companies like OpenAI, Meta, and Google have voiced concerns about the potential impact of such regulations on innovation. In response to these criticisms, Senator Wiener made several concessions in the latest version of the bill. Notably, the bill no longer proposes the creation of a new agency dedicated to AI safety. Instead, the responsibility for overseeing AI safety would fall under the existing California Government Operations Agency.
Additionally, the amendments clarify that companies would only be liable if their AI technologies caused actual harm or posed imminent threats to public safety. Earlier versions of the bill allowed for penalties even if no harm had occurred, which raised concerns among tech companies about the potential for overregulation.
“The new amendments reflect months of constructive dialogue with industry, startup, and academic stakeholders,” said Dan Hendrycks, a founder of the nonprofit Center for AI Safety in San Francisco, which played a role in drafting the bill. Despite these changes, not all concerns have been fully addressed. A Google spokesperson noted that the company’s previous concerns “still stand,” while Anthropic, another AI startup, indicated that it was still reviewing the amendments. OpenAI and Meta declined to comment on the revised bill.
Balancing Innovation and Safety
Senator Wiener remains optimistic that the bill strikes the right balance between fostering innovation and ensuring public safety. “We can advance both innovation and safety; the two are not mutually exclusive,” Wiener said in a statement on Thursday. He believes the amendments have addressed many of the industry’s concerns, paving the way for the bill’s passage.
The Democratic-majority Legislature is expected to pass the bill by the end of August. If it passes, the bill will be sent to Governor Gavin Newsom, who has not yet indicated whether he supports the legislation.
More Read
Broader Implications
If signed into law, SB 1047 would position California once again as a leader in tech regulation, following in the footsteps of its 2020 privacy law that curtailed the collection of user data and its 2022 child online safety law. However, opponents of the AI bill argue that it could stifle innovation, particularly by discouraging tech companies from sharing AI software code—known as open-source software—with other developers and businesses. Chris Nicholson, a partner at Page One Ventures, expressed concerns that the bill, even in its amended form, could “be a damper on AI development.”
The bill has also sparked political debate, particularly in San Francisco, a hub for AI startups. Several mayoral candidates have criticized the legislation, arguing that it threatens the city’s status as a leader in technology and innovation. “San Francisco’s reputation as a leader in technology and innovation is because we embrace new ideas and serve as a sanctuary for research and development, and SB 1047 clearly threatens our brand and leadership,” said Mark Farrell, a former interim mayor and current mayoral candidate.
Looking Forward
As the bill moves closer to a vote, it has become a flashpoint in the broader debate over how to regulate AI in a way that protects public safety without stifling innovation. While the amended version of SB 1047 has alleviated some concerns, critics like Lauren Wagner, an investor and researcher with experience at Google and Meta, argue that AI regulation should be handled at the federal level rather than by individual states.
With a vote expected by the end of August, the outcome of SB 1047 could have significant implications not just for California, but for AI regulation across the United States.
