The California State Assembly Appropriations Committee has recently approved a revised version of the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, spearheaded by State Senator Scott Wiener. Initially designed to prevent catastrophic AI failures, the bill has undergone significant changes following criticism from industry stakeholders and federal lawmakers. The amendments were introduced to address concerns that the original legislation could hinder innovation and drive AI companies away from California.
The revised bill removes a contentious provision that would have allowed the state attorney general to sue AI companies for negligence before a disaster occurred. Instead, the bill now permits the attorney general to request a halt to dangerous operations but reserves the right to initiate legal action only after a catastrophic event has taken place. Additionally, the new version eliminates the requirement for AI companies to submit safety test results under penalty of perjury. Developers are now expected to exercise “reasonable care” to ensure their models do not pose catastrophic risks, a reduction from the previous “reasonable assurance” standard.
The bill’s adjustments aim to strike a balance between ensuring AI safety and maintaining a supportive environment for technological innovation. Despite these changes, some critics, including notable tech figures and lawmakers, argue that the amendments still fall short in addressing the bill’s broader impact on the industry. Federal Democratic representatives, such as Reps. Zoe Lofgren and Nancy Pelosi, have expressed concerns that the bill could drive AI firms to relocate or discourage them from releasing models in California.
As the bill moves forward, it must pass the final vote in the California Assembly before being referred to the Senate. While supporters, including AI pioneers Geoffrey Hinton and Yoshua Bengio, view the revised legislation as a sensible approach to balancing innovation and safety, opposition remains strong. Venture capitalists and some legislators have called for the bill’s veto, arguing that it may not effectively protect against AI-related risks while potentially stifling scientific and technological advancement.
Reference:
