California contemplates distinctive safety regulations for AI companies, encountering resistance from tech firms

California lawmakers are weighing a groundbreaking bill that would mandate artificial intelligence companies to test their systems and implement safety measures to prevent potential future threats, such as disruptions to the state’s electric grid or aiding in the creation of chemical weapons. The vote is set for Tuesday.

The legislation aims to mitigate risks from AI advancements and faces strong opposition from tech giants like Meta and Google, who argue the focus should be on the misuse of AI rather than developers. The bill, authored by Democratic state Sen. Scott Wiener, targets extremely powerful AI models that could be developed in the future, specifically those requiring over $100 million in computing power to train. Currently, no AI models meet this threshold.

Sen. Wiener emphasized that the bill addresses potential future large-scale AI systems, not existing smaller models. Democratic Gov. Gavin Newsom, a proponent of AI adoption and regulation in California, has not commented on the bill but has cautioned against overregulation.

The proposal, backed by leading AI researchers, includes creating a state agency to oversee AI developers and establish best practices, with the state attorney general empowered to enforce compliance. However, a coalition of tech companies argues that the bill would hinder AI development, jeopardize open-source models, and cause regulatory fragmentation. Meta’s Rob Sherman expressed concerns that the bill would make the AI ecosystem less safe and drive companies out of California.

Tech companies prefer waiting for federal guidance, but supporters of the bill believe immediate action is necessary, drawing lessons from previous delays in regulating social media. Additionally, lawmakers are considering another measure to combat AI-driven discrimination in job and housing applications.