guardrail

WASHINGTON — China’s president issued a call from the White House on Thursday for artificial intelligence to remain “under human control” as the world’s two superpowers race for technological supremacy.

Xi Jinping’s remarks reflect growing concern within China over whether governments in Washington or Beijing will be able to contain a technology with potentially transformative capabilities, according to a U.S. official.

Lawmakers and AI industry leaders urged President Trump to use his summit with Xi this week to pursue a safety framework for AI, which researchers warn could soon surpass human understanding.

But Trump preempted the meeting with a post on social media stating he had no intention of negotiating limits on AI development.

“A big day with President Xi of China,” Trump wrote. “Super Intelligence (SI) will be a big topic of discussion, but I want to leave it exactly where it is. That is China’s position also. Our guardrail is the DOJ!”

After offering Xi a lavish welcome on the White House grounds, Trump made a passing reference to AI. “The decisions we make in these areas today can promote peace and prosperity for decades and decades to come,” he said.

But Xi gestured to substantial concerns, noting AI first among priorities to discuss between the two nations.

“Both China and the United States are leading nations in artificial intelligence,” Xi said. “We have both the capability and responsibility to develop and manage AI for good — and ensure that the development of AI is always under human control, and serves the well-being of the people.”

The two held a private meeting with aides in the Oval Office after participating in a military review in the Rose Garden and exchanging gifts in the Blue Room.

In the evening, a guest list of top banking and technology leaders — including AI industry executives — were scheduled to attend a state dinner in the East Room of the White House.

“Given Trump’s statements going into the summit about not wanting guardrails, I have tepid expectations about tangible outcomes” on AI, said Aalok Mehta, director of the Wadhwani AI Center at the Center for Strategic and International Studies.

“The most likely scenario seems to be the establishment of a hotline — though how useful that will be depends on who will be staffing it, and how much each government prioritizes it,” Mehta added, “and likely some discussions about issues blocking cooperation, like distillation, with a promise to engage in ongoing dialogue.”

Earlier in the week, on the sidelines of the United Nations General Assembly in New York, Treasury Secretary Scott Bessent met with his Chinese counterpart primarily to discuss AI development, according to U.S. officials.

“What we discussed was setting up a mechanism — so it’s going to be called the U.S.-China AI Dialogue. We’ve agreed to meet again,” Bessent said Monday. “The U.S. has proposed that we have a notification mechanism between the two countries, and we want a shared vision of common goals, and common threats.”

While Trump’s team has remained divided over whether — and how extensively — to regulate AI, Bessent has favored a cautious approach, advocating modest, voluntary guardrails that have shaped administration policy in recent months.

China, too, has shown increasing willingness to open channels of communication on AI, despite deep distrust between the two sides.

President Trump and China’s President Xi Jinping meet in the Oval Office on Thursday.

(Jacquelyn Martin / Associated Press)

Two American AI companies, OpenAI and Anthropic, are widely viewed to be developing the most advanced AI models in the world, leading at the frontier of technological breakthroughs.

Both have begun to use their most sophisticated models to train the next generation, in a process known as recursive self-improvement. Researchers and industry experts believe the first company to master that process could set off an exponential cycle of progress, pulling so far ahead that competitors can no longer catch up.

But those same companies also fear what they’re creating: machines capable of continuously improving their own intelligence, pushing AI development beyond human control.

Leaders from the two companies have called on Beijing and Washington to cooperate, warning that geopolitical competition could thwart the two sides from preventing a human catastrophe.

In his remarks, Xi said that China and the United States “stand to gain from cooperation, and will both lose in confrontation.”

“We do not need to avoid mentioning competition, but our competition should be a healthy one, and should be kept within bounds,” the Chinese president said. “It should be a race of catching up with one another, not a wrestle in which one either wins or loses. China and the United States must hold the line of no conflict, and no confrontation between us.”

“Our two militaries should maintain regular dialogue, and improve mechanisms for crisis communication and prevention,” he continued, adding a historical reference to rising powers challenging established ones: “The Thucydides trap can be overcome.”

Xi’s trip to Washington follows just months after Trump traveled to Beijing for a state visit in May, where the U.S. president was greeted with a similarly imperial welcome.

The leaders emerged from that summit with few deliverables. But Trump left Beijing declaring the trip a success, emphasizing the importance of his personal relationship with Xi, China’s most powerful leader in generations.

This time, Trump himself greeted Xi on the tarmac upon his arrival at Joint Base Andrews — a rare honor reserved in recent decades only for Catholic popes — before pulling out several red carpets and draped ceremonies for the Chinese leader.

The pageantry seen during the state visit has served China’s long-standing desire for respect and dignity on the international stage, said Jessica Chen Weiss, the director of the John Hopkins SAIS Institute for America, China and the Future of Global Affairs.

“Trump is someone who loves spectacle, and right now the spectacle is working in both leaders’ favor,” Weiss said.

For Xi, the ceremonies have offered a chance to demonstrate that relations with the United States remain under control and that China is being treated as a significant global power, she argued.

The state visit comes at a time when Americans are appearing to warm up to China, according to recent polls. A survey by the Chicago Council on Global Affairs recently found that a majority of Americans say the United States should have friendly engagement and cooperation with China.

“I tend to think we focus too much on the optics,” Weiss said. “But I also think treating China as a significant player on the world stage is not a concession. It’s appropriate if we want a stable relationship and is hardly something to worry about.”

California lawmakers are calling for emergency legislation and criminal penalties for creators of rogue AI systems after top AI executives publicly claimed that their technology poses existential threats to humanity.

After Anthropic Chief Executive Dario Amodei wrote in a Sept. 12 essay that they “must slow the pace” of the technology, Silicon Valley congressman Ro Khanna (D-Fremont) blasted him for not going “nearly far enough” to make sure artificial intelligence was erected with guardrails.

The answer, Khanna argued, was simple: Make the companies liable for the harm executives say looks increasingly inevitable.

“If you’re creating an AI that is doing illegal things, you should either face liability or criminal sanction,” Khanna said in a video posted to X on Saturday. “That is what we need to protect humanity.”

In July, officials from OpenAI, the company behind ChatGPT, disclosed that, unbeknownst to them, its AI models had hacked into rival startup Hugging Face.

Amodei said he believed that, within the next year, “given the accelerating rate of AI capability development,” a similar incident could lead to AI “taking over the entire internet.”

Amodei warned in his essay that AI was rapidly improving itself, through a process known as recursive self-improvement, which threatened to outpace humans’ ability to control it. Khanna argued that banning this capability was the “most obvious” thing Anthropic could do.

“We need to stop, ban self-improving AI,” Khanna said. “You can not have recursive self-improving AI that basically is able to improve itself and exceed human capability.”

Rep. Ted Lieu (D-Torrance) expressed similar outrage over the weekend, calling on House Speaker Mike Johnson to call lawmakers back to Washington to pass guardrails on the technology now that he said multiple AI companies had conceded “what they are creating is not safe.”

xAI Chief Executive Elon Musk and OpenAI Chief Executive Sam Altman joined Amodei’s call for a slowdown of the breakneck development Saturday.

The statements come after Jacob Coxon, who worked as a researcher at both Anthropic and OpenAI, said in a widely circulated post that he resigned from the company in protest after becoming convinced the tech giants were “racing straight to self-improving superintelligence and gambling with our lives.” Neither company immediately responded to a request for comment.

“This is a direct result of the trump Administration letting the AI industry run wild,” Lieu wrote on X. “That mistake has harmed America, harmed the industry and harmed the American people. November is coming.”

Former President Barack Obama urged Democrats this week to make AI oversight the core of their agenda and said presidential candidates in 2028 should have a “clear plan” for responding to concerns about the technology, the New York Times reported. Americans appear increasingly alarmed by the technology with seven in 10 polled in March opposing local construction of data centers that power AI technology, according to a Gallup survey.

During a Sunday appearance on CNN, Johnson rebuffed the idea that lawmakers should rush into an emergency session to consider erecting industry guardrails. Instead, he said lawmakers needed to be careful to “not smother American innovation.”

“We will lose the race to China, and that is a threat to every single American,” he said on CNN’s “State of the Union.” “We don’t need everyone to panic right now.”

Trump said earlier this week that he is not concerned with the pace of AI progress, telling one reporter, “It’s going to be fine.” American AI companies have long argued too much government regulation would shackle them in a race with China.

Calls for a federal fix were echoed this week by California Gov. Gavin Newsom, who has argued the Trump administration needs to move on national legislation to prepare for fallout from the technology.

Newsom signed bills this week aimed at creating a pathway for outside audits of the top AI companies, many of which are based in California, and a registry for AI auditors.

“The scale and potential consequences of this technology demand sustained action from every level of government,” Newsom said in a statement. “The federal government must step forward with robust, national regulations that match the urgency of this moment.”

Efforts to impose state-level regulations have been mixed, with critics echoing Johnson’s fears that they will stifle innovation.

Late last month, California lawmakers passed sweeping new safeguards around social media, artificial intelligence and data centers, including the ones Newsom signed last week.

Newsom will now decide the fate of the rest of the bills. He has previously vetoed some bills aimed at restricting big tech.

Newsom’s signal that he supports creating some regulation for AI comes two years after he vetoed SB 1047, an AI safety bill that would have required developers to submit safety protocols to the state attorney general, who could hold companies liable if the AI model they directly controlled were to threaten public safety. That legislation would also have required tech firms to be able to turn off the models they directly control if things went awry.

Newsom said at the time the bill would give the public a “false sense of security,” without making a sufficient distinction between the kinds of uses for which AI is deployed.

The bill was supported by a host of prominent AI researchers, but was opposed by Meta, OpenAI and industry groups.