US and China Announce AI Safety Protocol Plan After Trump-Xi Beijing Summit

--- headline: "US and China Announce AI Safety Protocol Plan After Trump-Xi Beijing Summit" slug: us-china-ai-safety-protocol-trump-xi category: policy story_number: "14" date: 2026-05-14 author: The Vault AI Staff tags: [us-china, ai-safety, ai-policy, geopolitics, trump, xi-jinping, bessent] ---

# US and China Announce AI Safety Protocol Plan After Trump-Xi Beijing Summit

The world's two largest artificial intelligence powers are preparing to sit down and write the rules of the road. Treasury Secretary Scott Bessent announced on Wednesday that the United States and China will establish a formal protocol governing AI safety best practices, a significant step toward bilateral governance of a technology that both nations view as central to their economic and military futures.

The announcement came during President Donald Trump's two-day visit to Beijing, a summit already loaded with agenda items spanning tariffs, rare earth minerals, semiconductor export controls, and Iran. But it was Bessent's remarks on AI, delivered in a pre-recorded interview with CNBC from the Chinese capital, that introduced the most concrete new deliverable of the trip.

"We're gonna set up a protocol in terms of how do we go forward with best practices for AI to make sure non-state actors don't get a hold of these models," Bessent said, framing the initiative squarely around proliferation risk rather than the broader question of how Washington and Beijing compete in the technology itself.

Two superpowers, one shared concern

Bessent was blunt about why the talks are happening now and on these terms. The United States holds a decisive advantage in frontier AI development, and that leverage, he argued, is precisely what makes cooperation possible.

"The reason we are able to have wholesome discussions with the Chinese on AI is because we are in the lead," Bessent told CNBC. "I do not think we would be having the same discussions if they were this far ahead of us."

That framing reveals the strategic calculus at work. Washington is not proposing to slow its own AI development or loosen the semiconductor export controls that have constrained Beijing's access to the most advanced chips. Instead, the protocol would focus on a narrower but universally compelling objective: preventing the most capable AI systems from being acquired or weaponized by terrorist organizations, criminal networks, or other non-state actors.

Bessent described the initiative as grounded in American standards, saying Washington would use "US best practices, US values" to chart the course forward "and then roll those out to the world." The language suggests the administration views the bilateral protocol not as a one-off agreement with Beijing but as a template for global AI governance, with the United States as its primary author.

The summit backdrop

The AI safety announcement did not emerge in a vacuum. Trump's Beijing visit unfolded against a backdrop of intensifying competition across multiple technology fronts. Washington has maintained strict controls on exports of advanced AI chips to China, though Reuters reported that roughly 10 Chinese companies have recently been cleared to purchase Nvidia's H200 processor, the second most powerful chip in the company's lineup. Nvidia CEO Jensen Huang was himself in Beijing during the summit, underscoring the commercial stakes.

Beijing, meanwhile, has been pressing Washington to roll back semiconductor restrictions and resume access to cutting-edge chip-making equipment. China has also wielded its own leverage by tightening export controls on rare earth minerals critical to the American automotive and aerospace sectors. The AI safety protocol, in that context, represents a rare patch of common ground in an otherwise adversarial technology relationship.

The bilateral dynamic has also shifted since China's DeepSeek rattled Silicon Valley earlier this year by demonstrating that competitive large language models could be built at a fraction of the cost assumed by Western labs. That breakthrough undermined the assumption that export controls alone could maintain a durable American lead, adding urgency to diplomatic channels that address AI risks directly.

What the protocol could look like

Details remain scarce. Bessent offered no timeline for when formal talks would begin, no indication of which agencies or ministries would lead the negotiations, and no specifics on what enforcement mechanisms might accompany any eventual agreement. He did describe the effort as focused on "the highest performance calculus where we can get the most innovation and the highest level of safety," suggesting the administration wants to avoid any framework that could be read as restraining American AI companies.

The most likely starting point, based on Bessent's comments, is a set of shared principles around model access controls, export screening for frontier AI systems, and information-sharing protocols when either government identifies attempts by non-state actors to acquire advanced capabilities. Whether Beijing would agree to transparency measures that could expose the reach of its own domestic AI ecosystem remains an open question.

Why this matters

A formal US-China AI safety protocol, even a preliminary one, would represent the first bilateral governance framework between the two nations specifically targeting artificial intelligence. Previous diplomatic efforts, including the November 2023 agreement between Biden and Xi to discuss AI risks, produced little in the way of concrete mechanisms. If the Trump administration follows through, the protocol could establish a precedent for how rival powers manage a technology whose capabilities are advancing faster than any regulatory apparatus can track.

The risks of inaction are not abstract. Frontier AI models are increasingly capable of assisting with tasks ranging from cyberattack planning to biological weapon design, and the barrier to accessing open-weight versions of near-frontier systems continues to fall. A protocol that even modestly improves coordination between Washington and Beijing on preventing the worst-case scenarios would address a gap that the global governance community has flagged repeatedly.

For now, the announcement is a commitment to talk, not an agreement. But in a bilateral relationship where technology has become the sharpest edge of competition, the decision to talk at all counts as news.