In a significant development in California, a piece of landmark legislation has passed a crucial vote in efforts to establish pioneering safety regulations for the largest artificial intelligence systems in the state. The proposal aims to minimize potential risks associated with AI by mandating companies to conduct tests on their models and publicly disclose their safety measures. This is to prevent the misuse of AI that could have severe consequences such as compromising the state’s electric grid or aiding in the development of chemical weapons, scenarios that experts believe could become possible with the rapid advancements in AI technology.
The bill, which is one of many being voted on during the final week of the session, narrowly cleared the Assembly vote on Wednesday and will now proceed to a final vote in the Senate before potentially reaching Governor Gavin Newsom’s desk for approval. Backers of the legislation assert that it would establish much-needed safety guidelines for large-scale AI models in the U.S., targeting systems that require over $100 million in data for training, a threshold which current AI models have not reached.
Despite its supporters, the proposal faced opposition from venture capital firms and tech giants such as OpenAI, Google, and Meta (parent company of Facebook and Instagram), who argue that federal regulations should handle safety guidelines for AI instead of the state level. Some critics of the bill, including a group of California House members and former House Speaker Nancy Pelosi, deemed it well-intentioned but misguided.
The Chamber of Progress, a Silicon Valley-funded group, criticized the bill, saying it is based more on science fiction depictions of AI rather than real-world considerations. However, the legislation garnered support from AI startup Anthropic, backed by Amazon and Google, after modifications were made based on the company’s input. Notable figures like Elon Musk, owner of social media platform X, have also lent their support to the bill.
The bill’s author, Democratic Senator Scott Wiener, emphasized that the legislation took a balanced approach towards AI innovation and safety. With concerns about the potential risks associated with powerful AI systems, Wiener stressed the importance of addressing these risks to prevent catastrophic misuse.
As AI technology continues to influence various aspects of society, California lawmakers have introduced numerous AI-related bills this year to promote public trust, combat algorithmic biases, and prohibit certain activities involving deepfakes. California, being home to a significant number of AI companies, has been at the forefront of AI adoption and might soon implement generative AI tools for addressing issues like highway congestion and road safety.
Governor Newsom, who had previously expressed concerns about overregulating AI, will have until the end of September to decide on signing the bill into law, vetoing it, or allowing it to become law without his signature.