California Governor Gavin Newsom vetoed a controversial artificial intelligence safety regulation bill on Sunday. The bill had previously received support from advocacy groups and Hollywood, while also drawing criticism from major Silicon Valley tech giants.

The vetoed Senate Bill 1047 (SB 1047) primarily targeted the model development stage, and originally planned to introduce whistleblower protections, strengthen safety requirements, and authorize the state Attorney General to bring civil lawsuits against non-compliant companies.

In his veto message to the California State Senate, Newsom said: "Despite good intentions, SB 1047 does not consider whether AI systems are deployed in high-risk environments, whether they involve critical decisions, or the use of sensitive data. Instead, the bill imposes strict standards even on the most basic functions—as long as the system is deployed by a large enterprise."
In his announcement on Sunday, the Governor stated that he is working with stakeholders in the field to develop more effective legislation.

Jennifer Everett, partner in the technology and privacy practice at Alston & Bird LLP, pointed out that frontier model developers most directly affected by the bill "may be able to breathe a sigh of relief."

Tech companies, including leaders from firms such as OpenAI, Meta, and Google, previously expressed regret over the bill's broad, "one-size-fits-all" language.

The bill has now been returned to the state legislature, where a two-thirds majority vote in both chambers could override Newsom's veto. However, the Center for AI and Digital Policy (CAIDP) noted in a Monday newsletter that veto overrides are extremely rare in California politics.

CAIDP stated: "This legislation, which we strongly supported, aimed to establish critical safety guardrails for large-scale AI models to prevent potential catastrophic harm." The organization called the veto a setback for AI safety advocates.

Various stakeholders have already begun looking ahead to next steps.

Peter Guagenti, President of Tabnine, said in an email to CIO Dive, a sister publication of CFO Dive: "Regulators need to hold big tech accountable, requiring genuine transparency in data usage, and refusing to accept opaque systems and rigid, unchangeable solutions. This may affect their operating costs, but it will build broader trust in AI and ultimately help us create a more dynamic and profitable ecosystem."

Despite the veto of SB 1047, Newsom has signed 17 other bills this month related to the deployment and development of generative AI, requiring watermarking of AI-generated content and combating AI-generated disinformation.

Everett said: "This will not be the end of AI regulation in California."

Top technology executives at financial services institutions, auto insurance companies, and consumer goods giants are advancing AI adoption plans that prioritize regulatory compliance.

Everett said: "The AI regulatory landscape across U.S. states will continue to evolve and will be quite complex."

Most enterprises are entering the fourth quarter with continued focus on AI, including deploying use cases and bringing together various stakeholders.

Everett said: "Companies should be cautious not to confine AI to a specific framework. These proposed laws emphasize that the use, deployment, or modification of AI systems requires more comprehensive participation from various stakeholders within the enterprise."

Growing regulatory concerns among technology leaders can be partly attributed to lagging best practices. According to a study released in August by the IBM Institute for Business Value, most C-suite technology executives admit their organizations lag in deploying responsible AI practices at scale.

Establishing meaningful governance structures will significantly impact whether companies can easily meet compliance requirements in the future. So far, common themes in regulatory provisions focus on transparency, ethical practices, and safety, distinguishing high-risk systems from other use cases.

As CIOs continue to assess the regulatory environment, one law that has begun to attract attention is Assembly Bill 2013 (AB 2013). This bill requires developers to publish documentation online about the datasets used in development, expanding transparency requirements.

Everett said: "AB 2013 is notable because its definition of developer is quite broad. It covers not only companies that produce AI systems, but also those that may make substantial modifications to AI systems, potentially including conversational product search, shopping assistants, or chatbot support."

The documentation requirements take effect on January 1, 2026, but apply to systems released after January 1, 2022, a timeframe that covers the public debut of ChatGPT and the subsequent surge in generative AI adoption.

California is not the only state where lawmakers are working on AI standards.

Colorado passed an AI-focused consumer protection law, approved by the Governor in May. Oregon, Montana, and Tennessee have also enacted AI-related legislation, and numerous states across the country have proposed related provisions.

Federal lawmakers in the United States are also working to advance AI legislative proposals, while in the European Union, the countdown to enforcement of its AI Act is underway.