United States of America

[US] California Governor vetoes AI safety bill

IRIS 2024-9:1/9

Maria Bustamante

European Audiovisual Observatory

On 7 February 2024, Senator Scott Wiener introduced the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act (SB 1047). Further information about this bill can be found in IRIS 2024-7:1/3. The Senate adopted it with a majority of 32 votes to 1 on 21 May 2024.

In response to industry advice, Wiener substantially amended the bill on 15 August 2024. The amendments included various clarifications and the removal of the proposal to create a Frontier Model Division and the penalty for perjury.

On 28 August, the bill was adopted by the State Assembly by 48 votes to 16. Then, following the amendments, it was once again voted on by the Senate and passed by 30 votes to 9.

However, the bill, which aims to mitigate the risks associated with AI and impose safety regulations on AI companies, was vetoed by Governor Gavin Newsom on 29 September.

The California Governor said he had decided to block the implementation of the regulation because the application of the bill and its consequences were unclear. “While well-intentioned, SB 1047 does not take into account whether an Al system is deployed in high-risk environments, involves critical decision-making or the use of sensitive data,” he wrote. “Instead, the bill applies stringent standards to even the most basic functions - so long as a large system deploys it. I do not believe this is the best approach to protecting the public from real threats posed by the technology.”

Zoe Lofgren and Nancy Pelosi, Democrat members of the US Congress, welcomed Gavin Newsom’s decision. In a statement, Zoe Lofgren said she thought the issue should be handled at federal level, both in the United States and abroad, in order to make it easier for companies that design and use AI systems to comply with the law.

Meanwhile, Daniel Castro, Vice President of the ITIF (Information Technology and Innovation Foundation), which promotes public technological innovation policies, issued a statement highlighting the other Californian AI-related bills signed by Governor Newsom. This other legislation mainly concerns deepfakes and digital likeness. The California Governor said he had approved 17 laws in 30 days in order to regulate generative AI.

The veto was criticised by leading Silicon Valley tech entrepreneurs, who fear that it could slow the pace of AI-related innovation and give a significant advantage to other countries in the development of future AI tools.