US China AI talks begin as superpowers agree to establish guardrails for powerful models

US China AI Discussions Launch After Beijing Summit

The United States and China have agreed to initiate formal discussions on AI governance, marking a significant diplomatic development following President Trump's summit with Chinese President Xi Jinping in Beijing 1

1

. Treasury Secretary Scott Bessent confirmed on Thursday that the world's "two AI superpowers are going to start talking," establishing protocols to prevent non-state actors from accessing powerful AI models 1

1

5

5

. This represents the first time the two countries will formally address AI issues during Trump's second term, as concerns mount that the technology could be weaponized by hackers and terrorists or spiral beyond human control.

China's foreign ministry spokesman Guo Jiakun confirmed that both heads of state "held constructive discussions on AI-related issues and agreed to launch an intergovernmental dialogue on artificial intelligence" 3

3

. President Trump told reporters on Air Force One that they discussed "possibly working together" on AI guardrails, though he provided limited specifics beyond mentioning potential areas including biological, nuclear, and cyber risks 4

4

.

AI Guardrails to Address Security Threats

The upcoming US and China AI discussions will focus heavily on establishing AI guardrails to manage emerging security threats. Bessent emphasized that establishing a protocol for best practices is critical "to make sure non-state actors don't get a hold of these models" 3

3

. The urgency has intensified following Anthropic's recent release of Mythos, a powerful cybersecurity model that can spot decades-old vulnerabilities, giving hackers the ability to easily target U.S. banks, government systems, or other software 4

4

. Anthropic limited Mythos' release to select technology firms, Wall Street banks, and government groups, demonstrating how the cyber risks of AI have moved from hypothetical to real.

Analysts identified autonomous AI weapons, cybersecurity threats, and AI-designed bioweapons as mutual concerns for both leaders at the Trump-Xi summit 3

3

. In 2024, Xi agreed with Biden that humans must remain in control of nuclear weapons decisions, establishing a precedent for cooperation on existential risks 3

3

. Sun Chenghao, a senior fellow at Tsinghua University's Center for International Security and Strategy, noted that "compared with 2024, the topics to be discussed this time might be broader," potentially including AI's impact on youth employment and other societal concerns 3

3

.

AI Supremacy Shapes Cooperation Terms

The fierce AI race between the superpowers continues to shape the parameters of cooperation. Bessent stated candidly that the United States is willing to engage on AI safety because "the Chinese are substantially behind us" in technological development 1

1

. "I do not think we would be having the same discussions if they were this far ahead of us. So we're going to put in U.S. best practices, U.S. values, on this, and then roll those out to the world," Bessent explained 1

1

. Experts suggest China's AI models may be a few months behind leading U.S. models, though the gap remains a major consideration for U.S. policymakers 1

1

.

Bessent emphasized that maintaining America's lead is "of utmost importance," adding that the U.S. must balance innovation with safety in AI development: "What we don't want to do is stifle innovation. So our responsibility is to come up with the highest performance calculus where we can get the most innovation and the highest level of safety" 5

5

. Officials and experts in both countries have argued they cannot slow technological development and risk losing to their rivals, creating a fundamental tension in cooperation efforts 1

1

.

Divergent AI Risk Perceptions Complicate Dialogue

A significant challenge for the intergovernmental dialogue stems from fundamentally different AI risk perceptions between the two nations. American experts generally highlight existential risks, such as the possibility of artificial general intelligence or super-intelligence exceeding human capabilities 1

1

. Chinese researchers and officials more often emphasize risks related to social stability and information control, including concerns about chatbots producing content that challenges China's leadership and policies 1

1

.

Despite these differences, researchers in both countries have identified shared concerns, particularly around AI being used to develop new biological weapons 1

1

. Guo emphasized that as major powers in the field, the countries should work together "to ensure that AI better serves the progress of human civilisation and the common well-being of the international community" 3

3

. Former President Biden's administration launched an official AI dialogue with China in 2023, but talks achieved limited success as U.S. experts focused on safety while China concentrated on U.S. export controls on AI chips 4

4

.

AI Chips Remain Contentious Issue

While AI governance took center stage, the contentious issue of AI chips lurked in the background of the Beijing summit. Nvidia CEO Jensen Huang joined Trump's entourage, hoping to secure a deal to sell Nvidia's H200 chip to China 2

2

. Trump cleared the way late last year for Nvidia to sell the more powerful H200 chips to China for a 25 percent cut of revenue, but Chinese firms have hesitated to purchase after the Chinese government encouraged them to support domestic chipmakers like Huawei 2

2

4

4

.

Trump initially told reporters the chip sale "didn't come up," but then contradicted himself, saying "it did come up and I think something could happen on that" 4

4

. The White House recently accused Chinese entities of "industrial-scale" efforts to steal U.S. technology, while Beijing blocked Meta's acquisition of a Chinese-founded AI agent tool, highlighting ongoing tensions 3

3

. Sun suggested keeping thorny issues like high-end chip exports separate from AI governance discussions "may help create a better atmosphere for talks between the two sides" 3

3

. White House officials indicated earlier that talks could include establishing a communications channel for discussing technology developments, though it remains unclear whether this was finalized at the summit 4

4

.