...
By John Feng and Didi Kirsten Tatlow
0
Share
Newsweek is a Trust Project member
See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.
China is brushing off warnings by U.S. company Anthropic that the development of artificial intelligence must slow down or lead to catastrophic consequences for humanity. That’s convenient for President Donald Trump, who believes AI risks are a big hoax.
“Hoax Buster” Trump says the United States is leading the AI race and must emerge victorious in order to reap its rewards: "Bigger than Oil, Gold, Diamonds, or even the Internet.” He said on Truth Social this week that his Chinese counterpart, Xi Jinping, “will be doing absolutely nothing to stand in the way of AI, or its future.”
All the signs suggest China is ordering its AI labs to advance in double time, but Beijing does worry about AI, just not for the same reasons that keep the Anthropic chief executive, Dario Amodei, up at night. While Amodei and co. fear a web takeover by rogue agents, the Chinese Communist Party sees AI as a possible threat to its regime stability, and its answer is to take full control of the capability—and quickly.
Read More on World
Anthropic has clashed with the Pentagon on moral grounds this year, but Amodei is a China hawk who knows how to keep the U.S. at the front of the tech race. When Trump hosts Xi in Washington, D.C., next week, the two men may unite on AI’s long-term benefit to their economies, but the U.S. president’s distrust of America’s AI industry leaders could be his biggest blind spot yet.
U.S. Big Tech Triggers Trump—and China
Over the weekend, Amodei argued in an essay that the U.S. should pace the frontier of AI or risk unintended events like loss of human control. Evan Hubinger, Anthropic’s head of alignment science, responsible for ensuring AI models comply with human-defined guidelines, said recently the probability that AI could kill all humans was greater than 10 percent.
Trump has found the industry’s caution hard to fathom, calling it a left-wing conspiracy to hobble America. Slowing down, he said, would bankrupt U.S. big tech firms, or worse—cede the AI race to China. In his mind, that would mean a failure to lead a soon-to-be trillion dollar industry.
...
China’s nationalist state news outlets balked at Amodei’s manifesto too, but for different reasons. Their knee-jerk reaction was over the Anthropic CEO’s recommendations to ban Beijing’s top labs from accessing U.S. AI software and hardware, and to let America alone decide the speed of AI development. That can only happen if the U.S. stays No. 1.
Amodei’s thinking has support from rivals Sam Altman of OpenAI and Elon Musk of SpaceXAI, whom Chinese state media have collectively dubbed the “U.S. tech right.” They want to “curb China's AI development through technological barriers and regulatory monopolies, uphold Washington's monopolistic hegemony in cutting-edge technology, and exclude China from the global AI governance system,” the nationalist Global Times tabloid said.
The China Daily newspaper concluded: “The proposed coordination among the three companies sounds rather like a club whose membership rules have been drafted before the guest list is announced. A global AI-safety framework that excludes China is not quite global. It is a US AI elites' arrangement with an altisonant name.”
Beijing holds a particular disdain for Anthropic and its whistleblowing. Amodei’s company revealed in a September 10 report last week that China’s biggest AI labs were using its Claude agents without authorization to train their own models, including by rerouting queries from Chinese users handling highly sensitive data—a potential embarrassment for the Chinese government.
China Wants To Make AI Safe—for Itself
AI may turbocharge the Chinese economy’s shift from a goods manufacturer to a major tech power, but to capture global markets with its AI services and infrastructure for data and power, Beijing must take risks—and there are serious risks.
The world took notice this week when China’s intelligence chief Chen Yixin, who is the minister of state security, said the country would speed up the development of AI, integrate it into everything and simultaneously ensure that the CCP controls it politically and ideologically.
The CCP doesn’t believe in slowing down AI development, but it does believe in AI risks, especially to itself. Hostile foreign forces could misuse AI tools such as cheap generative images to manipulate public opinion, he said, among other warnings. In short, AI threatens the party’s usually tight hold on information and by extension the legitimacy of its rule.
The solution to the AI threat is to “adhere to the party's control of the internet and the party's control of data,” Chen wrote. That language means China’s main spy agency “is going to be used to bolster that industry and to protect it,” Peter Mattis, president of the Jamestown Foundation, a Washington, D.C., think tank, told Newsweek.
“They’re essentially saying: ‘We now consider that to be part of our job.' We may see them go on a broader offensive across their tech stack to do this,” Mattis predicted.
What’s Left To Talk About?
China’s president is scheduled to arrive in the U.S. in a week’s time and many observers expect some sort of AI safety dialogue to take place, although the framework remains unclear.
Both Trump and Xi want to accelerate their AI buildout and neither appears ready to accept limits. Xi in particular would not accept restrictions proposed by the U.S., its only rival in the AI race, according to Ryan Hass, a senior policy analyst at the Brookings Institution think tank.
“Beijing believes deeply the US is determined to hold China's AI progress back, especially with export controls. Beijing is unwilling to accept a subordinate position to US on AI,” Hass said on X. “Beijing is concerned about potential impacts of AI on regime security. But Beijing isn't going to follow the US lead in addressing those risks.”
Treasury Secretary Scott Bessent told Axios on Wednesday that the U.S. is “open to discussions on avoiding shared risks and avoiding bifurcation of our two systems.” AI researchers will find the comment promising but perhaps also too late, given the respective memberships of the U.S.-led Pax Silica and the China-led World Artificial Intelligence Cooperation Organization—two competing AI blocs.
OpenAI CEO Altman, who will be at Trump's state dinner for Xi, said recently that the American and Chinese leaders could win the Nobel Peace Prize if they agree on AI limits. Trump and Xi may not have an incentive to slow down, but they could both benefit by agreeing in principle where to stop.
The same Anthropic report that called out China last week also found that U.S.-built models were being exploited in scarily creative ways, including to harm U.S. interests.
Iran and its allies have used Claude to collect open-source intelligence on U.S. Navy personnel and to produce targeting recommendations, while the Tehran-backed Houthi rebels in Yemen have used Anthropic’s agents to develop and test guided rockets and ballistic missiles, the report said.
Beijing can surely see that its own enemies could do the same.
A version of this article was published in Newsweek's China Decoded newsletter on Wednesday, September 16. Sign up here.
View 0 Comments
Submit Correction
View Editorial & AI Guidelines