California lawmakers have voted in favor of a controversial bill, known as Senate Bill 1047, which requires companies that develop or modify powerful forms of artificial intelligence to test their models for potential harm to society. The bill states that companies spending a significant amount of money on AI, either through training or modifications, must evaluate the model's capacity to enable cybersecurity or infrastructure attacks, as well as the development of harmful weaponry. The bill has received opposition from companies like Google and Meta, as well as AI experts who argue that compliance costs and legal concerns may hinder innovation and the release of open source AI tools. On the other hand, whistleblowers and advocates for AI regulation support the bill, claiming that AI tools pose significant risks and that government regulation has been inadequate.

The bill's author, Senator Wiener, stated that its purpose is to uphold safety testing standards already agreed upon with President Biden and other global leaders. Despite concerns that the bill may lead to businesses leaving the state, Wiener dismissed this argument, citing previous instances where such predictions did not materialize following the adoption of other tech regulations. The bill went through several amendments before passing, with proponents arguing for the prevention of potential AI disasters and the need for stronger legislation. Additionally, the California Legislature is also considering other AI-related laws, including takedowns of deepfakes, AI usage guidelines for schools, combatting automated discrimination, and measures to protect voters from deceptive deepfakes.