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

The United States and China will start discussing A.I. safety, Bessent says.

The United States and China will discuss guardrails on artificial intelligence, including establishing a protocol for keeping powerful A.I. models out of the hands of nonstate actors, Treasury Secretary Scott Bessent said on Thursday. Mr. Bessent, who was speaking from Beijing in an interview with CNBC, did not give more details, including when these discussions would take place. But Xi Jinping, China's leader, and President Trump had been expected to discuss A.I. during their summit in the Chinese capital. If these talks happen, it would be the first time the two countries formally take up the issue during Mr. Trump's second term. The capabilities and usage of A.I. have grown rapidly, and so have concerns that this technology could be weaponized by hackers and terrorists, or spiral out of human control. "The two A.I. superpowers are going to start talking," Mr. Bessent said. "We're going to set up a protocol in terms of, how do we go forward with best practices for A.I. to make sure nonstate actors don't get ahold of these models." Still, Mr. Bessent made clear that the fierce competition between the United States and China for supremacy in A.I. -- which has been a major hurdle to cooperation on safety -- remained front of mind for U.S. policymakers. Officials and experts in both countries have argued that they cannot slow technological development and risk losing out to their rivals. Mr. Bessent said that the United States was willing to cooperate with China on A.I. safety because "the Chinese are substantially behind us" in terms of the technology's development. "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," Mr. Bessent said. Experts have suggested that China's A.I. models may be a few months behind the leading U.S. models. Another hurdle to the United States and China working together on A.I. safety is that they have generally focused on different potential threats. American experts have generally highlighted existential risks, such as the possibility of artificial general intelligence, or super-intelligence that exceeds that of humans. Chinese researchers and officials have more often highlighted risks related to social stability and information control, such as the possibility of chatbots producing content that challenges China's leadership and policies. Still, researchers in both countries have highlighted some shared risks, such as the possibility of A.I. being used to develop new biological weapons.

How A.I. Was the Elephant in the Room at the Trump-Xi Summit

President Donald Trump's entourage of tech and business leaders should have placed artificial intelligence at the center of his highly-anticipated summit with Chinese President Xi Jinping in Beijing this week. But the leaders appeared to focus more on limited questions of trade, without reaching any agreement on the future of A.I. A lot seemed to be on the table for A.I. The talks, which concluded on Friday, took place against the backdrop of a global A.I. race that has been dominated by the two superpowers. Democratic lawmakers have raised the alarm on allowing Chinese firms to buy A.I. chips from the U.S., and the White House has in recent days accused China of mass A.I. theft. At the same time, Nvidia CEO Jensen Huang, who joined Trump on the trip at the last minute, was hoping to secure a deal to sell China Nvidia's H200 chip, which has not been delivered to China despite U.S. approval for its sale. Chinese firms have dragged their feet in purchasing the chips after the Chinese government encouraged them to turn to domestic chipmakers like Huawei.

Beijing says China, US should work together to promote AI governance

Beijing (AFP) - China and the United States "should work together to promote the development and governance of AI", Chinese foreign ministry spokesman Guo Jiakun said on Tuesday. Cooperation on artificial intelligence was discussed by US President Donald Trump and China's Xi Jinping at talks in Beijing last week, both sides say, despite their countries' fierce rivalry over the fast-evolving technology. "The two heads of state held constructive discussions on AI-related issues and agreed to launch an intergovernmental dialogue on artificial intelligence," Guo told a news briefing, confirming previous remarks by US Treasury Secretary Scott Bessent. As major powers in the field, the countries should also work together "to ensure that AI better serves the progress of human civilisation and the common well-being of the international community", Guo added. Analysts said before the summit that fears over autonomous AI weapons, cybersecurity and the threat of new AI-designed bioweapons were mutual concerns for Xi and Trump. In 2024, Xi agreed with Trump's predecessor Joe Biden that humans must remain in control of the decision to fire nuclear weapons. But with China set on narrowing the United States' lead in the strategic sector, until now little further cooperation has followed. The White House recently accused Chinese entities of "industrial-scale" efforts to steal US technology, while Beijing blocked the acquisition of a Chinese-founded AI agent tool by tech giant Meta. But at the same time, the AI cybersecurity threat has been highlighted by Mythos, a powerful new model that US startup Anthropic withheld from public release to stop it from being exploited by hackers. Bessent told CNBC on Thursday that Washington and Beijing would set up a "protocol" on the path forward on AI, particularly "to make sure non-state actors don't get a hold of these models". The world's "two AI superpowers are going to start talking", Bessent said. While details on what will be discussed are so far scarce, Sun Chenghao, a senior fellow at Tsinghua University's Center for International Security and Strategy, told AFP that "compared with 2024, the topics to be discussed this time might be broader". "The two sides could share some best practices and exchange experiences on how to address and manage" AI's impact on society, for example on youth employment, he said. "Even though China and the United States are in competition in the field of AI, the impact of AI technologies on the entire world -- and on all kinds of actors, whether states, societies, or businesses -- is extremely significant." However, keeping thorny issues such as the export of high-end US chips that can train and power AI systems for separate meetings "may help create a better atmosphere for talks between the two sides", Sun added.

Trump says he discussed AI guardrails with Xi

President Trump said Friday he discussed "working together" on artificial intelligence guardrails with Chinese President Xi Jinping as policymakers in Washington debate on how manage the risks of the emerging technology. "We talked about possibly working together for guardrails" on AI, Trump told reporters on Air Force One following his talks with Xi in China. When asked what kind of guardrails, the president provided little specifics, stating the "guardrails that we talk about all the time." The topic of AI was largely expected to come up in this week's summit to Beijing, which is racing to stay competitive with the U.S. when it comes to the development of the technology. While the Trump administration, along with Chinese leaders, have largely pushed forward with a pro-innovation, light-touch regulation landscape, new models like Anthropic's Mythos are forcing Washington to rethink its strategy. "AI is fantastic," Trump said Friday, "So many things can happen in terms of health, medicine and operations, everything...military." "So many things can happen, but it's also got some drawbacks and we're talking about....we probably will, we're going to work together," he added. The president did not explain which types of risks, but when asked by a reporter if this will mean biological, nuclear or cyber, Trump shook his head yes and said, "Could be, yeah." Treasury Secretary Scott Bessent told CNBC on Thursday that the U.S. can hold AI talks because the country is "in the lead" in the race with China. "The two AI superpowers are going to start talking. We're going to set up a protocol in terms of how do we go forward with best practices for AI to make sure nonstate actors don't get a hold of these models," Bessent said. White House officials suggested earlier this week that talks about AI could include ways to open up a communications channel for the two nations to discuss technology developments. "Like in many areas of intense focus for the US and China, it's good to have a channel of communication," one U.S. official said Sunday, adding, "What that channel of communication looks like and its formality ... is yet to be determined. But we want to take this opportunity with the leaders meeting to open up a conversation and see if we should establish a channel of communication on AI matters." It is not clear whether Trump's talks with Xi included this communications channel, which would be a first for the Trump administration. Former President Biden's administration launched an official AI dialogue with China in 2023. But talks had limited success as U.S. technical experts at the time were more concerned about safety, while China was focused on U.S. export controls on AI chips. The meeting comes about a month since Anthropic released its cybersecurity model Mythos to a limited group of technology firms, Wall Street banks, and government groups. The model, according to Anthropic, is moving the cyber risks of AI from the hypothetical to real life. The AI firm says Mythos can spot decades-old vulnerabilities, giving hackers the ability to easily target U.S. banks, government or other software in breaches. Mythos' release rattled the Trump administration, which typically prioritized the U.S.'s competitive standing over AI safety risks, though it doesn't appear there is a consensus now on how to address concerns. Notably, Trump told reporters Friday the sale of Nvidia's advanced computer chips to China did not come up, as some predicted it might. "It didn't come up," the president said, adding China has not approved of the purchase of Nvidia's H200 chips "because they chose not to, they want to develop their own." He then said, "but it did come up and I think something could happen on that." Nvidia CEO Jensen Huang was one of more than a dozen technology and business leaders who accompanied Trump in China. Huang, for his part, has pushed back on U.S. export controls on chips, arguing they helped China accelerate the development of its own technology. Trump cleared the way late last year for Nvidia to sell its more powerful H200 chips to China for a 25 percent cut of revenue.

US, China are discussing AI guardrails to safeguard most powerful models, Bessent says

WASHINGTON, May 14 (Reuters) - U.S. and Chinese delegations will discuss artificial intelligence guardrails at their Beijing summit and will set up a protocol for best practices to keep non-state actors from getting the most powerful AI models. Bessent told CNBC in a pre-recorded interview on Thursday that it was "of utmost importance" that the U.S. maintain its lead over China in AI, adding that this is why Beijing is interested in discussing guardrails. "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," Bessent said. (Reporting by David Lawder; editing by Susan Heavey)