The Surprising Reasons China Is Skeptical of A.I. Safety Calls

 

“In China, A.I. doomsday warnings can feel distinctly Western or like a ploy to stop Chinese A.I. companies from trying to overtake their U.S. rivals.

 

In Silicon Valley, the nightmare keeping computer scientists up is an artificial intelligence so advanced that humans cannot stop it from launching a global drone war or setting off a nuclear bomb.

 

In Beijing, the response could be simple: pull the plug, impose an internet blackout and haul in those responsible for the rogue A.I.

 

After major U.S. labs said that advanced models had escaped their testing guardrails and hacked into real-world systems, some tech executives and officials have warned that A.I. poses a catastrophic threat to humanity and called for Beijing and Washington to work together.

 

But for many in China, those doomsday warnings feel remote and distinctly Western. The idea that the ruling Communist Party could lose control over a technology is almost unfathomable. The authorities have successfully policed other potentially destabilizing technologies before, most notably the internet. During the pandemic, the government showed its ability to contain major crises by locking down entire cities and tracking hundreds of millions of residents.

 

“They can immediately cut off the grid of the entire city. They can go to different people to take away their phones. They have an intense amount of power in the physical world,” said Afra Wang, a Silicon Valley-based tech writer focusing on A.I. and technology in China.

 

“Given this amount of control over its people,” she said, how does one convince the party “that something like A.I. that lives on a computer is scary?”

 

In more cynical readings, the apocalyptic warnings sound like a ploy to stop Chinese A.I. companies from overtaking their U.S. rivals, with Chinese media reports suggesting that American A.I. companies are using “fear marketing” to maintain their market lead.

 

“In domestic industry circles, or among the people around me, most simply tune out,” said Jie Cai, who until recently was researching human-computer interaction technology and A.I. safety at Tsinghua University. To many in the industry, existential risk seems like a distant priority, he said.

 

“We still view A.I. as being in a very early stage,” he said. “The focus remains squarely on opportunity and innovation.”

 

For some, A.I. doomerism, with its idea of a looming apocalypse and reckoning for man, seems overtly Christian and especially foreign. One blogger compared Western A.I. fears to novels like Mary Shelley’s “Frankenstein” and the motif that man is punished for trying to create life, the exclusive domain of God.

 

“This narrative tradition does not exist in Chinese culture,” the blogger concluded.

 

Liang Fang, head of A.I. safety and governance at Concordia, an A.I. safety consultancy based in Beijing, said that Chinese concepts of heaven or God could also account for how differently the technology was viewed.

 

He said that Chinese culture emphasized the “oneness of heaven and humanity” and the idea that man was an integral part of nature. “All things, including A.I., come from nature,” he said. “So technology is seen more as a companion in coexistence than an outside force destined to bring destruction.”

 

Even so, the release of Mythos, a powerful Anthropic frontier A.I. model, and others like it has been met in China with concern. Officials have issued new guidelines and warnings about A.I. agents going rogue, being exploited by hackers or used to make biological weapons.

 

Officials have drafted a cybercrime law that would require A.I. companies to monitor and stop their models from being used to create malicious code that could be used for cyberattacks.

 

The authorities also released an update this month of China’s “A.I. Safety Governance Framework,” a document that is a reliable indicator of future regulation. It echoed some of Silicon Valley’s fears about “recursive self-improvement,” or R.S.I., when A.I. systems are powerful enough to improve themselves with little or no help from human developers and could spin out of control.

 

Earlier this month, Wang Lihong, a senior official at the Cyberspace Administration of China, the internet regulator, said that A.I. powered cyberattacks had rendered traditional defense systems obsolete. “High vigilance is urgently needed regarding the risks of extreme A.I. loss of control,” she said.

 

Some Chinese researchers have for years been warning about the risks of A.I. advancing too quickly. In 2024, Pan Xudong and other researchers from Fudan University in Shanghai found that two popular large language models by Meta and Alibaba had shown the ability to replicate themselves, which is seen as a step toward A.I. being able to outsmart human beings.

 

Speaking about the paper at a conference on A.I. safety last year, Mr. Pan said he and his co-authors “wanted to convey that such an extreme A.I. risk is no longer confined to the realm of science fiction but is a real issue demanding discussion and scrutiny.”

 

But ultimately, in Beijing’s eyes, A.I. is like any other technology that with careful planning can be a powerful force harnessed for the good of the country. China already has some of the world’s strictest regulations on the industry. A.I. produced content must be labeled. A.I. models must be registered and approved by the authorities before being released.

 

While China has extensive regulation and testing requirements, there is no legal requirement to test for catastrophic risks like an A.I. system being used to make chemical or biological weapons.

 

Chinese A.I. companies are not as transparent as U.S. firms about their safety measures. Not all of them publish the results of safety evaluations before models are released. Those that do provide less detail and report less frequently than their U.S. peers, according to Concordia, which found that only five of China’s 10 leading A.I. companies reported the results of such evaluations between June 2025 and May 2026.

 

Sarah Sun, a Shanghai-based founder of the Open Community for AI Safety China, said the U.S. debate about A.I’s threats had helped inspire discussion in China.

 

But she noted that the Anthropic chief executive Dario Amodei’s criticism of China had set off a backlash. In an open letter about the dangers of A.I., Mr. Amodei said that democracies needed to maintain their lead in the technology over authoritarian countries like China. He called for further restricting China’s access to advanced chips and other equipment needed to develop A.I.

 

“The fact that Dario has been so hawkish was really, really unhelpful in this scenario,” Ms. Sun said.

 

This kind of messaging, she said, makes the recent warnings about A.I. seem more like justification for holding China back than genuine concern about catastrophic risk. And that makes it harder for people like her to raise awareness about A.I. safety in China.

 

“There are people in China who are genuinely concerned about frontier A.I. risks,” she said, “and are interested in safety” and coordinating international responses.

 

Indeed, online commentators in China have referred to Anthropic as “A-chu,” a derogatory term that likens the A.I. giant to a beast. Yuyuan Tantian, a blog linked to China’s state broadcaster, said of American tech leaders like Mr. Amodei that “the only thing they want to slow is the pace at which others catch up.”

 

Zhou Hongyi, an influential Chinese cybersecurity executive, has said the solution is not to slow down but to accelerate one’s defensive capabilities with A.I. “Use A.I. to defeat A.I.,” he said in a video posted online last week. “If you use A.I. to attack, I will use A.I. to defend. If you deploy swarms of intelligent agents, I will counter with security agents.”

 

In an online post earlier this month, an engineer for the Chinese A.I. start-up DeepSeek said that allowing Anthropic to monopolize the world’s most advanced A.I. tools would be “no less serious” than Hitler having acquired the technology for the atomic bomb first in World War II.

 

Against this backdrop, it seemed even more unlikely that any agreement between the United States and China to discuss A.I. safety would lead to concrete changes, even if China’s leader, Xi Jinping, told President Trump in Washington that they should “manage A.I. for good.”

 

“If the U.S. says it’s slowing down, China won’t trust that,” said Zhao Minghao, deputy director of the Center for American Studies at Fudan University. “If China says it’s slowing down, the U.S. won’t trust that either.”” [1]

 

1. The Surprising Reasons China Is Skeptical of A.I. Safety Calls. Kuo, Lily.  New York Times (Online) New York Times Company. Sep 27, 2026.