US and China Discuss New AI Safety Alert System Ahead of Trump-Xi Summit

US and China Discuss New AI Safety Alert System Ahead of Trump-Xi Summit

Imagine two countries racing to build the world’s most powerful AI systems.

Now imagine those same countries agreeing that if something goes seriously wrong, they should at least tell each other. That is the conversation happening between the United States and China.

Ahead of a planned meeting between U.S. President Donald Trump and Chinese President Xi Jinping in Washington on September 24, senior officials from both countries have discussed a possible new system for reporting major AI incidents that could threaten national security. And the timing is interesting.

AI is getting more powerful. The competition between Washington and Beijing is getting sharper. And concerns about what advanced AI could do if something goes wrong are growing.

A Possible “AI Alert” Between the US and China

The proposal came up during talks in New York between U.S. Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng.

According to Bessent, the U.S. proposed a new “notification mechanism” for AI incidents that could have national-security consequences.

The idea is relatively simple: if a serious AI-related incident happens, the two countries could have a way to notify each other rather than discovering it through rumors, intelligence reports, or social media.

Bessent said the goal is to move from an opaque relationship toward greater transparency between what he described as the world's No. 1 and No. 2 AI powers.

It would not mean the U.S. and China suddenly become AI partners.

Far from it. They are still competing intensely for technological leadership.

But AI creates a strange situation: the countries can compete with each other while still facing some of the same risks.

Why This Matters Now

The conversation comes at a time when researchers and AI companies are raising increasingly serious concerns about what advanced AI systems could enable.

Those concerns include sophisticated cyberattacks, misuse involving biological research, attacks on critical infrastructure, and the possibility of increasingly autonomous AI systems making consequential decisions.

Just days ago, U.S. and Chinese security experts also discussed safeguards for extreme AI scenarios, including keeping humans in control of consequential cyber operations and creating communication channels for incidents involving autonomous AI.

That makes the latest government-level discussion particularly notable.

The question is no longer simply:

“Who will build the most powerful AI?”

It is increasingly becoming:

“What happens if something goes seriously wrong?”

But There Is Still a Huge Trust Problem

This is where things get complicated. The U.S. and China are not simply friendly competitors.

They are engaged in a much broader technological and geopolitical rivalry, including disputes over advanced chips, AI models, technology transfers, and access to critical technologies.

Washington has accused Chinese AI developers of extracting capabilities from advanced U.S. models. Beijing has rejected those accusations.

At the same time, China is rapidly narrowing the AI gap with the U.S., making the technology race even more important to both governments. So building trust around AI safety will not be easy.

A notification system sounds simple. Getting two competing superpowers to trust the information being shared is another matter entirely.

Trump and Xi Are Now Set to Take Over the Conversation

The discussions between Bessent and He were part of the groundwork for Trump's planned September 24 meeting with Xi in Washington.

AI is expected to be among the issues discussed.

The two leaders have already discussed AI cooperation during their May meeting in Beijing, so this week's summit could move the conversation further. But expectations should be realistic.

A full U.S.-China AI safety agreement would be a much bigger step than simply establishing a mechanism for communication.

For now, the focus appears to be on something more basic:

keeping the two sides talking when an AI-related crisis crosses national-security lines.

The AI Race Is Changing

For years, the U.S.-China AI competition was mostly framed around chips, computing power, research, and who could build the smartest models.

Now there is another layer.

Safety.

If AI systems become capable of launching cyber operations, discovering vulnerabilities, assisting biological research, or interacting with critical infrastructure, an incident involving one country's systems could potentially affect the other. That makes communication increasingly important even between countries that do not fully trust each other.

The proposed notification mechanism does not solve the bigger AI safety problem.

It does, however, point to something important:

The AI race may be too consequential for countries to compete without also creating ways to manage the risks they share. And with Trump and Xi preparing to meet in Washington, the world is about to see whether that idea can move beyond discussion.