Bill Gates said Sunday that getting countries to agree on rules for artificial intelligence will be harder than the nuclear arms talks of the Cold War, and urged the United States to set the example, NBC News reported. Speaking to Kristen Welker on “Meet the Press,” the Microsoft co-founder said monitoring AI systems would not undercut America’s position against China.

What Happened

Asked whether a global framework for AI was achievable, Gates compared it with U.S.-Soviet arms control and judged the AI problem the more difficult of the two. He argued the U.S. should lead anyway, citing aviation safety, protection of the ozone layer and nuclear weapons as areas where Washington acted as what he called a fair actor rather than pressing its own advantages.

Welker raised President Donald Trump’s opposition to federal AI guardrails. Gates said he expected the president to reach a consensus after hearing from him and others, and argued that putting monitoring and safeguards in place would not handicap the country in its competition with China.

RelatedU.S. Begins Selecting Leaders for Program Enlisting Firms Against Foreign Cybercriminals

Gates said the most dangerous near-term risk is people with bad intentions using AI. He warned that small groups equipped with AI could carry out cyberattacks or bioterror plots that once required the resources of the largest nations, and listed fraud among the threats.

He also said a kill switch alone would not be sufficient. To curb harmful behavior, authorities need insight into what AI systems are doing and records of their activity, he said.

Why It Matters

Gates drew a line between two arguments that are often merged. Whether to slow AI development is a separate debate, he said, while deciding to build in safeguards and monitoring is a choice that can be made regardless. That framing puts him on the side of oversight without endorsing the calls for a slowdown that have come from inside the industry.

The interview came as the White House continues to resist new federal AI rules. NBC noted that Trump has said he wants superintelligence left exactly where it is and described that as China’s position too, and that Rep. Ro Khanna, a California Democrat, has proposed U.S.-China working groups of technical experts to ban recursive, self-improving AI models and set up verifiable monitoring.

After Trump’s meeting with Xi Jinping, the White House announced a dialogue with China on AI risks and benefits, with follow-up exchanges expected by November, according to NBC.

The Foundation’s AI Work

Gates described the Gates Foundation’s own AI efforts, which focus on languages other than English. He said error rates in English are roughly a tenth of those in African languages, and that the foundation is working with 60 AI companies to broaden the languages their systems handle reliably. The goal, he said, is that within five years every language spoken by at least a million people will approach English-level quality on these platforms.

He also addressed his past association with Jeffrey Epstein, calling it regrettable and saying Epstein had promised to raise money for global health that never materialized.

What to Watch

The U.S.-China AI dialogue and the exchanges expected by November are the first formal test of whether the two governments will discuss monitoring at all. Gates’ prediction that the administration will come around to safeguards runs against its current stance, and any shift would likely show up first in how Washington approaches those talks.