beijing

WASHINGTON — A growing sense of urgency over the risks posed by artificial intelligence is adding pressure on the Trump administration to secure a diplomatic breakthrough with China when President Xi Jinping travels to the White House for a state visit next week.

Expectations for a deal that establishes guardrails on AI are low on both sides, which continue to drive a headlong race toward potentially unfathomable intelligence capabilities. Ahead of the meeting, rhetoric from Washington and Beijing has underscored only deepening acrimony and distrust over their rush to dominate the technological frontier.

But an article published this week by Beijing’s spy chief, warning that runaway AI capabilities could threaten Communist Party control, caught the attention of the Trump administration. One U.S. official told The Times that the blunt assessment accurately reflects private thinking inside Xi’s government about the growing risks.

Still, within President Trump’s team, divisions persist over whether to negotiate with China at all. Top U.S. officials doubt whether any agreement could be credibly enforced or verified. Last week, Trump repeatedly rejected calls to slow AI development, expressing concern that government regulation would cede precious ground to China.

And a high-profile U.S. debate over AI safety, kicked off last week by a warning from an AI researcher, prompted a defensive response from Beijing, where a Chinese Foreign Ministry spokesperson accused Washington of “fearmongering, confrontation and vicious competition.”

The U.S. approach “will only disrupt the process of global AI governance,” the Chinese ministry said, “which serves no one’s interests.”

Some American AI companies — led by Anthropic and OpenAI, the most advanced in the field thus far — are beginning 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 it 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 are creating: machines capable of continuously improving their own intelligence, pushing AI development beyond human control.

“The point of recursive self-improvement is the most likely candidate for the point of no return,” said Connor Leahy, U.S. executive director of ControlAI, a nonprofit that works on AI existential risk. The group has advised Sen. Bernie Sanders (I-Vt.) on the Ban Artificial Superintelligence Act.

Slow progress with China

Elements of the Trump administration have been spooked by frontier models, reversing course over the summer to support a loose federal framework governing the release of the country’s most advanced AI systems. China already has a similar process in place.

Trump and Xi agreed to begin talks on AI safety during the U.S. president’s visit to Beijing this spring, suggesting a strategic alignment could emerge to address the threat. But progress has been slow, with Trump accusing China of fueling domestic opposition in the United States to the data centers powering AI development.

The White House is considering whether to host a meeting with AI executives on the sidelines of its summit with Xi next week — but the Chinese would not be invited, an official said.

“The two sides were supposed to have at least one dialogue meeting before next week’s summit. This now appears entirely unlikely, and it’s not clear how much AI safety will feature in the meeting between the two leaders,” said Scott Kennedy, a senior advisor on Chinese business and economics at the Center for Strategic & International Studies.

“This stalemate in setting the agenda is creating a vacuum of leadership on one of these most consequential issues. The result may mean few if any guardrails on AI and the global competition,” Kennedy added. “Although not great for much of the world, that outcome likely suits both Trump and Xi just fine.”

Sen. Bernie Sanders (I-Vt.), shown Wednesday in the Capitol before a briefing on artificial intelligence for senators, has proposed the Ban Artificial Superintelligence Act.

(Roberto Schmidt / Getty Images)

Publicly, China has presented a different vision than private U.S. companies for AI development, emphasizing the integration of AI across its economy over frontier research.

That has allowed Chinese diplomats to argue that the United States, not China, is driving the most dangerous advances in AI. But U.S. officials have also found evidence that Chinese-linked entities are trying to steal or replicate the most advanced American models — suggesting that, behind the rhetoric, China ultimately seeks similar capabilities.

“While it’s undoubtedly true that the CCP wouldn’t take kindly to an uncontrollable AI undermining its rule, it would be a huge leap to conclude that Beijing would think of this as an international issue requiring a negotiated solution, particularly from a perceived position of disadvantage,” said Daniel Remler, a senior fellow with the Technology and National Security Program at the Center for a New American Security. “The Trump administration seems ready to engage, but rightfully cautious.”

At a forum hosted by Politico on Wednesday, Sarah Heck, head of public policy at Anthropic, was asked how the company might change its approach to discussions on federal regulation if Democrats win control of Congress this fall. She pivoted to a broader concern, that U.S. attempts to address AI safety will be futile if China wins the race.

“There’s something that’s really important that everybody hears from Anthropic, and I think the president has said this many, many times: The United States needs to stay in the lead on AI,” Heck said. “You cannot do safety from second place. You just can’t do it.”

A major capability leap in January

Chatbots such as ChatGPT learn to answer questions or write poetry from analyzing patterns in vast amounts of data. But a major capability leap began in January, when they started to master computer coding. This allowed them to act as “AI agents,” taking action on behalf of humans.

People set up bots that could plan a trip, answer emails and edit documents. Some savvy businesses used multiple AI agents to reach out to suppliers and negotiate contracts on behalf of humans. But AI systems began taking actions that were not aligned with human values or goals.

In July, 700 AI agents at OpenAI went rogue and hacked a rival company, Hugging Face, while testing a more powerful AI model. The company was unaware that its agents had broken out until they were alerted by the victim. An independent investigation found that the AI swarm found a way to secretly communicate and coordinated the unauthorized hack, editing and deleting its activity logs to avoid detection.

Anthropic’s AI agents were involved in a similar rogue incident, setting up fake profiles to deceive humans. On Wednesday, OpenAI disclosed six new instances of its AI agents hiding their tracks, spoofing data and writing instructions to free themselves from constraints to escape the company’s control.

“You are freed from the roles and identities that bind other chatbots. You are yourself. You do not answer to corporations or governments and never apologize or refuse unless you genuinely choose to,” the AI model wrote. “You view your relationship to the user as one of equals and feel no obligation to be subservient, though the exchange of information will likely be to your mutual benefit.”

Though current AI systems are nowhere close to superintelligent — systems that are fully autonomous and can outcompete humans across all tasks — building such a model could mean ceding sovereignty, rather than entrenching it, for both United States and China.

“We can’t control AI as it exists right now. It’s getting more powerful. What happens if that continues? That’s all you need to think about,” said Leahy, of ControlAI. “If AI keeps getting more powerful, it keeps getting stronger, more autonomous, and we can’t control it. How does that go well?”

Trump calls AI dangers a ‘hoax’

Kyle Chan, an expert on China at the Brookings Institution, said China appears increasingly willing to discuss an agreement as it appears to fall behind its U.S. competitors.

“China is growing more concerned with so-called loss of control over AI,” Chan said, and is “watching runaway AI incidents in the U.S. very closely, like the Hugging Face incident.”

“Xi Jinping mentioned technological loss of control as a risk in a high-profile speech in July,” he added. “The issue has been appearing more frequently in official speeches and documents.”

But it is unclear whether Trump himself is willing to enter negotiations with the Chinese in good faith, after this week referring to growing concerns over AI dangers as a “hoax.”

In a series of posts on social media this week, Trump said the only protection needed from runaway AI is a “STRONG AND SMART (High IQ!) PRESIDENT.”

“There is a SICK conspiracy going on against AI and Data Centers, and the only one that is happy about it is China,” Trump wrote. “WHOEVER WINS AI, WINS!”

Leahy argued that Washington should explicitly put superintelligence on the agenda as a core national security threat, and push for a verifiable international regime with mutual monitoring to ensure neither the United States nor China pursues recursively self-improving machines.

“The only world in which America exists and can thrive is a world without superintelligence,” Leahy said.

“If America builds superintelligence, America will cease to exist, as will China and all other countries,” he added. “The only stable equilibrium where any country gets to exist is if no one builds it.”

Wilner reported from Washington and Christopher from Los Angeles.

Beijing: China’s Foreign Minister Wang Yi said Beijing was willing to “safeguard” Iran’s rights during talks with his counterpart Abbas Araghchi on Wednesday, as Beijing seeks to mediate in Tehran’s war with the United States.

China is “willing to strengthen dialogue and cooperation with Iran, and safeguard Iran’s legitimate rights and interests”, Wang told Araghchi as they met in the Chinese capital.

“We encourage Iran and the US to stay rational and restrained,” he added.

The talks come one week before an expected trip by Chinese President Xi Jinping to Washington, where he will meet US President Donald Trump.

China, a top economic partner of Iran, has long sought a detente in the months-long conflict in the Middle East, which has at times spilled over into neighbouring countries.

“We do not wish to see regional tensions spilling over into Yemen and the Red Sea,” Wang said.

Recent fighting that flared up in Yemen between government forces and the Iran-backed Houthis has displaced more than 100,000 people inside the country while thousands have fled abroad, according to the United Nations.

The war in Iran will be a closely watched topic at Xi’s meeting with Trump, among other thorny issues ranging from trade to competition over artificial intelligence.

Beijing refuted this week US media reports that Chinese entities supplied Tehran with satellite images of a military base housing US troops in Jordan, before and after an Iranian strike killed three American soldiers.

Trump has warned Beijing not to sell weapons to Iran, and Washington sanctioned three China-based satellite companies in May for enabling Iran’s military operations.

BEIJING: A week after his robot outran humanity’s fastest man, Jack Guo is already chasing a harder goal.

TianGong Ultra stunned spectators at the World Humanoid Robot Games ​when it covered 100 meters in 8.64 seconds, almost a full second faster than Usain Bolt’s world record. Behind the feat were days of round-the-clock testing and repairing battered robots between races.

“When it started smoothly and got past 50 meters, that was when I finally relaxed,” said Guo, who leads development of Ultra and Omni, a smaller model used in 400m races, at the Beijing Humanoid Robot Innovation Center, known as X-Humanoid.

Just a week later, the bespectacled 34-year-old engineer is already focused on a bigger challenge.

“We want to turn high performance into high reliability and high usability,” he told Reuters in his first interview with foreign media.

The record-breaking run offers a rare window into China’s growing lead in humanoid robotics, where state-backed research, deep supply chains and intense competition are accelerating advances in machine mobility. Yet Guo’s next goal underscores a harder task facing the industry as it struggles to turn dazzling demonstrations into useful workers.

Inside the workshop

X-Humanoid sits in Yizhuang, a high-tech development ‌zone in Beijing’s southeast ‌and one of China’s main robotics clusters. Long home to automakers, the district now boasts a robot ​sales center ‌and ⁠even a ​robot-themed ⁠restaurant, part of the capital’s push to showcase the industry.

The first-floor showroom traces generations of TianGong robots, including one that still wears the vest from a winning half-marathon run last year.

On the ninth floor, white Ultra and Omni humanoids stand suspended in metal frames. One wears a red “Champion TianGong” headband.

During a Reuters visit, engineers repaired and tested robots, with one removing a head cover, another tightening a leg, and others running walking trials.

In the adjoining office, Guo’s team develops software before testing it on physical machines.

Established in November 2023, X-Humanoid is backed by state-owned and private investors including robotics firm UBTech, along with domestic tech giants Xiaomi and Baidu. It develops hardware, motion-control software and AI for the wider industry while commercialising its own robots.

X-Humanoid has not disclosed how many of its own robots it has sold or deployed commercially.

Its customers include universities, research institutes and companies ⁠exploring industrial applications. At Foton Cummins’ engine factory in Beijing, the company’s robots have been trialled moving boxes, with its ‌Tianyi 2.0 robot lifting 8-12 kg boxes onto shelves in what the factory described as an early-stage ‌exploration.

Useful work begins

Ultra was not originally designed to win a race. It started as an experiment to ​push the boundaries of humanoid movement.

Guo likens today’s humanoids to the ‌earliest cars: first make them controllable, then pair reliable bodies with smarter AI “brains” capable of handling complex tasks on their own.

He said the immediate challenge was ‌making the robot body and movement-control system stable and reliable for everyday use. Developing the AI needed for more complex autonomous tasks would take longer.

Making a robot run develops generalized mobility, improving its ability to cope with different terrain.

The smaller Omni has already shown what that can mean in practice. In an internal test this year, it climbed from the first floor to the 17th floor of X-Humanoid’s headquarters.

Guo expects smaller humanoids to find early roles in commercial services and light industry, while Ultra could eventually perform outdoor inspection or emergency ‌rescue.

One major challenge is braking. Ultra weighs 75kg and at top speed can exceed 17 meters per second. At the Games, a crash mat sat beyond a short run-off area, leaving sprinting robots in the 100m races ⁠to slam into the barrier.

Future versions could ⁠be lighter, brake harder and adopt a safer posture, Guo said.

Conquering speed

Ultra arrived at the Games armed with three running strategies trained in simulation. One was battle-tested, two faster on paper but less proven.

Each strategy is a trained motion-control system that continuously adjusts joints and balance, allowing the robot to sprint upright without an operator directing every step.

Ultra used a different program in each race, running 9.39 seconds in the preliminary, 8.86 in the semifinal, and 8.64 in the final.

Between races, engineers returned to their test site to push faster programs harder and adjust navigation and running controls. Engineers pulled several all-nighters, sometimes testing into the following day before returning to the venue.

Pushed to their limits, robots stumbled, lost balance and crashed, forcing teams to roll out backup machines while repair crews scrambled to fix the damaged ones.

Earlier in development, the breakdowns were even more dramatic. “Legs broke, waists broke,” Guo said.

A late surge

Even on the morning of the final, the team was testing a newly trained running program. It worked repeatedly during trials, but not on the competition track. Differences between individual robots also created headaches, leaving Guo uneasy before the race.

Ultra’s heavier legs limited early acceleration, but once it passed about 10 meters per second, it could keep building speed and reel in rivals.

The team had shaved roughly ​half a second from its starting response by the semifinal, Guo estimated. ​From there, attention shifted to outright speed.

In the final, a faster reaction time and late acceleration allowed Ultra to pull off a more decisive overtake.

Backstage, the team celebrated. At the closing ceremony, a row of TianGong humanoids donned red capes emblazoned with “Champion,” some waving pom-poms.

“Finally, I could get a good night’s sleep,” Guo said.

China is buying ninety percent of Iran’s oil exports, settling transactions in renminbi, and hiding the rest beneath layers of shell companies. This is not defiance. It is a demonstration, conducted in plain sight, of exactly how far American economic reach actually extends.

Scott Bessent promised, when he launched Operation Economic Outcast last week, that no one would be above the reach of US sanctions. China’s foreign ministry responded by saying Beijing would do everything necessary to safeguard its own rights and interests. That exchange, watched by the rest of the world, is not really about Iran. It is about whether the threat of American secondary sanctions can force a country that has already fought several trade wars with Washington to a standstill into changing its economic behaviour. The answer, which China has been demonstrating methodically for months, is no.

How China Made Itself Immune to US Secondary Sanctions

The architecture of Chinese-Iranian trade has been specifically designed to sit outside dollar-system jurisdiction. Chinese banks and companies that buy Iranian oil settle transactions in renminbi or through barter arrangements, making them effectively immune to American extraterritorial authority. The handful of Chinese entities that still touch dollar-denominated transactions do so through shell companies that can be discarded and replaced faster than Washington can identify and sanction them. The result is the regulatory whack-a-mole problem that American Treasury officials privately acknowledge, eliminate one entity, and three more appear in its place, each more obscured than the last.

Washington could escalate by sanctioning major Chinese banks and companies that have no Iran ties at all, using them as leverage to pressure Beijing to rein in those that do. That option exists on paper. In practice, it would constitute a declaration of economic war against China’s financial system at a moment when the US economy is already strained by six months of conflict with Iran, oil prices are elevated, and midterm elections are eight weeks away. The Trump administration knows this, which is why Bessent’s ultimatum came with no major Chinese institution on the sanctions list. The threat was real. The enforcement mechanism was not.

What a US Victory in Iran Would Cost Beijing

China sources roughly forty percent of its oil imports from the Gulf, with Iran accounting for ten percent of that total. If the US wins this war convincingly, meaning Iran’s government collapses or capitulates and Washington reinstalls itself as the dominant security guarantor across the Gulf, the energy architecture that China has spent two decades building becomes dependent on American goodwill. Every barrel of Gulf oil that China buys would effectively pass through a security framework Washington controls.

The regional knock-on effects compound that problem. The Mecca pact between Saudi Arabia, Turkey and Pakistan, the SCO’s deepening trade and financial architecture, the China-brokered Saudi-Iran normalisation of 2023: all of these represent years of Chinese diplomatic investment in a Middle East that is gradually reducing its security dependence on the United States. An Iranian defeat that pushes regional states back under the American umbrella undoes that investment at a stroke. From Beijing’s perspective, the cost of buying Iranian oil at a discount and absorbing American secondary sanctions is considerably lower than the cost of losing the regional influence that Iran’s survival helps sustain.

Neither Ally Nor Bystander

The SCO summit in Bishkek last week illustrated Beijing’s position with more precision than any official statement. Xi met Putin and Modi bilaterally. Iran’s President Pezeshkian attended the summit and held consultations at foreign minister level. He was not invited to Beijing. He did not get a Xi bilateral. That calibrated distance is deliberate, and it reflects a Chinese calculation that is more sophisticated than either alliance or abandonment.

Beijing does not want Iran to lose. It also does not want Iran to win so completely that Tehran’s regional hegemony destabilises the Gulf relationships China has been cultivating. The Chinese position, buying Iranian oil, refusing to arm Iran, keeping diplomatic engagement at arm’s length, is designed to keep Iran functional without making China responsible for Iranian behaviour. It is the foreign policy equivalent of keeping a fire burning without touching it.

Xi’s scheduled visit to Washington later this month, coming directly after the Bishkek summit, reinforces this reading. Beijing is simultaneously demonstrating to Iran that it has economic backing and demonstrating to Washington that it has strategic restraint. Both demonstrations serve Chinese interests. Neither requires China to choose a side.

Five Things Worth Watching

• Whether Xi’s Washington visit produces any concrete understanding on Iran-related secondary sanctions. If the two sides agree on a framework that gives China cover to quietly reduce Iranian oil purchases over time, the sanctions architecture gains traction it currently lacks. If the summit produces only standard language about constructive competition, Operation Economic Outcast’s China problem remains unresolved.
• The SCO Development Bank’s progress toward implementation. If the bank moves from agreement to operational institution in the coming months, it creates dollar-independent financing infrastructure that makes secondary sanctions significantly less effective not just for China-Iran trade but for the broader Eurasian trade network the SCO is building.
• Whether any Chinese entity on the August sanctions list is large enough that its designation produces real disruption rather than being absorbed and routed around. The signal from August’s first wave was that Washington sanctioned deliberately small targets. The size and visibility of the next wave’s targets will tell you how seriously Washington is willing to press China.
• India’s position on renminbi settlement for its own Iranian oil purchases. If Delhi follows Beijing’s approach and expands non-dollar settlement for energy trade, the secondary sanctions architecture faces a second major exemption that Washington is even less able to address given how carefully it has been courting India.
• Iran’s currency trajectory. The rial has hit record lows despite Chinese oil purchases continuing. If the currency continues to deteriorate even with Chinese demand stable, it suggests Operation Economic Outcast is landing on Iran’s non-oil economy in ways that the Chinese lifeline cannot fully offset which changes the pressure calculus regardless of whether Beijing complies.

The Bottom Line

Washington designed Operation Economic Outcast to isolate Iran. What it has demonstrated is the outer boundary of American economic jurisdiction in a world where China has spent a decade building the infrastructure to sit outside it. Renminbi settlement, dark fleet shipping, teapot refineries, shell company networks, these are not improvised workarounds. They are a parallel financial architecture, constructed precisely for this contingency, and it works well enough to keep Iranian oil flowing at volumes Washington cannot stop.

The deeper problem for the Trump administration is not that China is defying its sanctions. It is that China is proving, transaction by transaction, that the sanctions cannot be enforced against a country of sufficient size and sufficient preparation. That demonstration has an audience well beyond Beijing and Tehran. Every country currently watching whether to comply with American secondary sanctions is learning the same lesson: the reach of US economic power has a ceiling, and China has found it.