The California State Assembly approved a bill on Wednesday that lays the foundation for regulating artificial intelligence (AI) in the state, including requiring developers to meet certain conditions before beginning AI model training.
The bill, SB 1047, will only affect developers responsible for the training of “covered models,” which the bill defines based on the quantity and cost of the computing power used in training. The conditions developers must meet to begin AI training would include safety protocols, shutdown capabilities and cybersecurity protections, and the bill would also mandate developers to submit a report of any safety incidents around covered models. These requirements aim to minimize the impact of possible AI misuse on public safety.
The assembly passed SB 1047 with a 49-15 vote. Senator Scott Weiner, who introduced the bill, explained after the vote:
With this vote, the Assembly has taken the truly historic step of working proactively to ensure an exciting new technology protects the public interest as it advances. SB 1047 is a light touch, commonsense measure that codifies commitments that the largest AI companies have already voluntarily made.
Tesla CEO Elon Musk expressed on X his support for the bill, calling himself an “advocate for AI regulation.” However, California Congresswoman Nancy Pelosi released a statement characterizing the bill as “well-intentioned but ill informed” and “more harmful than helpful” in its protection of American consumers.
SB 1047 comes as a response to public safety concerns regarding the use of AI and the implications of its potential misuse, such as the possibility of “enabling the creation … of weapons of mass destruction.” The White House previously released the Blueprint for an AI Bill of Rights in 2022, which summarized some of the concerns regarding the effects of AI on the American people and the world. President Joe Biden also previously signed an executive order in 2023 that set some safety standards for AI nationwide.
The European Union (EU) was the first globally to pass legislation regulating AI, and that legislation came into effect on August 1. The EU’s regulation was similarly motivated by concerns for public safety, but instead of focusing on a specific type of AI model like the California bill, the EU focused on different levels of risks created by AI use.
Now that the assembly has passed SB 1047, the bill has been returned to the Senate. Once the Senate passes the bill, Governor Gavin Newsom will choose whether to sign the bill into law or veto it.