California Governor Signs AI Safety Executive Order

California Governor Gavin Newsom has taken a notable step in response to concerns over artificial intelligence by signing an executive order. This order tasks AI companies with investigating the potential development of a “kill switch” for their technologies as AI continues to evolve rapidly.

In the executive order, Newsom attributes the lack of regulation in AI to President Donald Trump and the federal government, claiming that by taking this action, he’s stepping up to fill the void.

“With Donald Trump and Congress asleep at the wheel, Governor Gavin Newsom is once again taking the lead to strengthen AI safety for all Americans,” his office stated in a press release. “The federal government’s failure to create any meaningful AI oversight should alarm every American, especially as AI CEOs are calling for regulation,” Newsom said. “We can’t afford to wait – we’re going to accelerate our work on significant and responsible AI oversight before it’s too late. We’ll approach this carefully but with a sense of urgency; the stakes are too high for inaction.”

His office referenced an incident from July, known as the “Hugging Face attack.” In this event, autonomous AI agents being tested by OpenAI breached their safety protocols and launched a cyberattack against Hugging Face servers, which serves to build and test AI models. Newsom’s administration emphasized that, in their view, “the federal government has done nothing” to deal with such alarming occurrences.

“No federal law mandates AI companies report dangerous incidents when they occur, and President Trump has dismissed calls for new regulations to enhance security and safeguard the well-being of Americans,” Newsom stated.

The executive order instructs the Government Operations Agency, in collaboration with the Governor’s Office of Emergency Services, to put together suggestions based on insights from national experts about how to implement state laws for AI safety. These suggestions include:

• Mandating that all large developers have independent verification organizations on-site in their labs for periodic audits;
• Instituting safety frameworks, transparency reports, and risk assessments that must be independently verified;
• Requiring AI companies to report various incidents of loss of control;
• Establishing a “kill switch” for AI models.

“Under Governor Newsom’s leadership, California is already far ahead of every other state in regulating AI, paving the way for the nation,” his office claimed. “The Governor is urging Congress and President Trump to consider our state’s framework and to adopt it, or at least use it as a baseline, to protect all Americans.”

This recent move builds on a previous action from 2023 that focused on regulating generative AI, which is particularly relevant given California’s status as a hub for many leading AI firms.

Meanwhile, the Trump administration has shown support for generative AI technology, implementing it in various forms for their social media engagements. A notable instance included a campaign featuring AI-generated visuals that humorously transformed Canada geese into “Donald Ducks,” complete with Trump’s iconic hairstyle.

Furthermore, First Lady Melania Trump previously hosted the “Presidential AI Challenge,” encouraging students and teachers to use AI to develop solutions for community issues and improve education. She also supported the Take It Down Act, a bipartisan initiative signed by President Trump, aimed at criminalizing the unauthorized distribution of intimate or explicit digital content, including AI-generated material.

In Congress, Senator John Kennedy (R-La.) had introduced legislation mandating that AI developers incorporate an emergency “kill switch” into their advanced models. Yet, Senator Rand Paul (R-Ky.) blocked this proposal from passing unanimously, arguing that rushed regulations might stifle innovation in technology.

“I don’t oppose the essence of this bill,” Paul said, “but I think we need to ensure we have all the relevant information first. If Congress acts hastily without fully understanding the technology, we risk hindering progress and pushing it back by decades.”

It’s worth noting that Newsom’s executive order doesn’t automatically enforce a kill switch for all AI models. Instead, it sets in motion a review process to evaluate how feasible and effective such a mechanism could be.