California Safe and Secure Innovation for Frontier Artificial Intelligence Models Act - S.B. 1047

Summary

SB1047, or the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, would require that a developer, before beginning to initially train a covered model, comply with various requirements, including implementing the capability to promptly enact a full shutdown, and implementing a written and separate safety and security protocol. It would also prohibit a developer from using a covered model for a purpose not exclusively related to the training or reasonable evaluation of it or make the covered model available for commercial or foreseeably public use if there is an unreasonable risk that it will cause or materially enable a critical harm. Additionally, developers would be required to perform audits via a third-party, independent auditor, which would then be provided to the attorney general’s office upon request. The attorney general would also be authorized to bring a civil action against developers who violate these requirements.

Updates

August 15, 2024. Advanced to the Assembly floor with amendments after industry engagement. Major amendments include replacing criminal penalties for perjury with civil penalties; scrapping a proposed new state regulatory body, formerly called the Frontier Model Division; narrowing pre-harm enforcement by cutting the attorney general's ability to seek civil penalties; adjusting the threshold to protect startups ability to "fine-tune" open sourced models; and adjusted the legal standard for developers from "reasonable assurance" to "reasonable care."

August 28, 2024. Passed the Assembly 48-16.

August 29, 2024. Passed the Senate 30-9. Waiting on approval or veto from Governor Gavin Newsom.