On Sunday, California Governor Gavin Newsom vetoed a bill that many thought would serve as a blueprint for AI legislation in the United States. If signed into law, SB 1047 would have imposed restrictions on the nation’s top AI companies running large AI systems, like establishing a state entity overseeing some AI development and mandating safety testing for the big AI model systems.
In a letter, Newsom described his reasoning for vetoing the bill, suggesting that the legislation may give the public a “false sense of security” since it only focused on large AI models. He also cited concerns that the legislation could curb AI innovation.
As it stands right now, federal regulation of AI systems has been slow to develop. But the Biden administration unveiled a sweeping executive action in October 2023 laying out some core tenets and actionable items, in addition to securing voluntary commitments from more than a dozen companies around safe development of AI systems.
RELATED STORY | Scripps News got deepfaked to see how AI could impact elections
To find out more about how the White House has been implementing the executive order and tackling this issue, the Scripps News Disinformation Desk sat down with Ben Buchanan, the White House Special Advisor on AI. While he usually serves as a professor at Georgetown University's Walsh School of Foreign Service, Buchanan went on leave more than three years ago to help the Biden administration better understand this issue that both excites and terrifies people.
“The president wants to set up a system in which AI systems are safe and secure before they're released to the public,” Buchanan said. “What really strikes me is he focuses on the technology's effect on people. He wants to know what does it mean for workers, what does it mean for consumers, what does it mean for ordinary American families.”
The Biden administration wants to strike a balance with AI policies that both protect privacy and workers and civil rights, while also balancing innovation. Buchanan said one of the ways the Biden administration is looking to mitigate the risks of AI, especially when it comes to disinformation generated by the systems, is by pushing for more labeling of AI content.
“There's a broader disinformation conversation and this is one part of it. But I think, what can AI do in a disinformation context? It can make things appear maybe more real than they are with deepfakes and the like,” Buchanan said. “And this is why we push so hard on watermarking and transparency, because that I think is the best way to mitigate that risk. And the core principle here is Americans should know when they're seeing a video that's generated by an AI system, and we need to build out a technical ecosystem that enables that in a seamless way for Americans.”
RELATED STORY | How to tell the difference between a deepfake video and a real one
From a national security perspective, Buchanan believes the United States’ most direct competition on this issue is with China.
“I think on national security across the board, our deepest competition is with China, and that's true in AI as well," he said. "It doesn't mean we're not willing to have conversations with China that are collaborative on AI, but I think it's fair to say that China is a primary competitor.”
Buchanan asserts that the United States has worked hard to establish a leadership position on AI – including from designing the critical chips to now trying to make those chips in the U.S. to assembling the large data centers that power complicated AI systems. And it’s also about retaining top human talent to make it all happen.
When asked about how he views the risks of AI – whether he’s more focused on long-term risks like killer robots that could be developed, or more concerned about short term risks like housing opportunities that are affected by an algorithm, Buchanan offers a sports analogy.
“AI policy is like a decathlon,” he said. “There's a lot of different events, and the United States can't skip any of them. We have to compete. We have to excel across the board.”
Buchanan also stated that he believes the U.S. energy grid can support artificial intelligence. According to a report released by the International Energy Agency at the beginning of this year, AI, along with data centers and cryptocurrency, consumed 2% of the global electricity demand in 2022. That is expected to double in 2026.
“We're going to build more clean power here in the United States,” Buchanan said. “We can handle the load right now. By definition, we have. The United States is the hub for AI training and operations for the world. But we know we’re going to have to build a lot more clean energy to keep that status.”
RELATED STORY | Arizona's secretary of state trains election officials on threats from AI, deepfakes
Buchanan also mentioned that the Biden administration recently convened a meeting between Energy Secretary Jennifer Granholm, leading AI companies, leading datacenter companies, and leading utility companies at the White House to discuss this very issue.
With the increasingly polarized political environment in the United States, the intelligence community recently warned that China, Russia, and Iran are using artificial intelligence increasingly to try to impact the upcoming election. Buchanan said he will let the intelligence community speak for itself, but added, “This I think reaffirms why we're talking about the importance of labeling AI content, labeling non-AI content, in cases where that's appropriate. So, will this sway the election? I doubt it. But we are trying to put as much information out to the public so that they are informed.”
Last week in his final speech to the United Nations General Assembly meeting, President Biden dedicated several minutes to addressing AI, a move that only further highlighted the pressing nature of this technology. At one point, Biden stated, "Artificial intelligence is going to change our ways of life, our ways of war and our ways of war."
Scripps News pressed Buchanan on what exactly the president meant with that last phrase.
“I think there's no doubt that AI offers the potential for a lot of capability in the military intelligence context,” Buchanan said. “The executive order tasks the development of a national security memorandum, basically guidance to the Department of Defense and the intelligence community about how to use this technology. The first ever document of this kind for AI, and that document is not out yet, but when it is out, that document will give very clear guidance about how AI changes the Department of Defense's strategy and crucially, how it doesn't. And there are some things that are really important, that retain within human control that we use AI in a way that aligns with our values."
