China’s state-run Global Times has published a sharply worded editorial accusing Anthropic CEO Dario Amodei of using AI safety language as cover for an effort to contain China’s technological rise, describing his recent call to slow down frontier AI development as a “Cold War playbook” dressed up as a rational safety proposal.
The editorial, published under the outlet’s Opinion/Viewpoint section, was written in response to Amodei’s essay “We Must Pace the Frontier,” which he published over the weekend. In it, Amodei argues that the pace of AI capability development needs to be deliberately slowed, warning that the industry’s ability to understand and control increasingly powerful models is being outrun by how fast those models are improving. The essay quickly drew endorsements from some of Amodei’s most prominent industry rivals, including OpenAI’s Sam Altman and Tesla and xAI’s Elon Musk, an alignment that is is notable mainly because of how publicly the three executives typically clash.
Global Times, however, reads the essay very differently. Its editorial argues that beneath the “rational statement” framing focused on global AI safety lies a set of “containment provisions targeting China,” and that Amodei’s proposal amounts to a strategy of choking off China’s AI progress first, only to later pursue “conditional negotiations” with Beijing to slow down once a technological gap has been secured.
Three measures Global Times objects to
The Chinese outlet’s editorial zeroes in on three specific elements of Amodei’s framework that it characterizes as containment measures rather than safety measures: a strict ban on selling advanced AI chips or semiconductor manufacturing equipment to China, a crackdown on so-called “model distillation,” and tighter security at AI labs to prevent the theft of model weights.
Amodei is explicit that pacing among democracies has a ceiling: it can only go as far as the lead the US currently holds over Chinese AI development, and he backs continued restrictions on chip exports to China, tougher enforcement against unauthorized distillation of frontier models, and stronger security to prevent model weight theft. These are close to word-for-word the same three measures Global Times is objecting to, though the two accounts diverge sharply on intent — Amodei frames them as buying safe time to pace responsibly, while Global Times frames them as the real payload of the essay disguised as a safety initiative.
The China angle isn’t new to Amodei’s public writing. Previously, he has made comments he made calling it “crazy” that the US was giving China access to even non-frontline chips, and data showing Anthropic talks about regulation and AI safety far more than OpenAI does, a pattern that piece noted critics see as convenient overlap between national security language and Anthropic’s own competitive interests.
“Hypocritical and short-sighted”
The Global Times editorial goes further than disputing individual policy proposals, arguing that the entire framing of a “silent AI Cold War” is hypocritical and short-sighted. It says Amodei’s real purpose is to use technological barriers and regulatory monopolies to curb China’s AI development, preserve Washington’s dominance in cutting-edge technology, and push China out of the global AI governance system altogether — comparing the effort to what it calls Washington’s “Pax Silica” initiative.
The editorial also takes issue with the idea that China’s AI capabilities can be meaningfully excluded from the global ecosystem at this point, calling China one of the largest and most open suppliers in global AI and describing any attempt to sever global AI supply and innovation chains as one that would primarily raise the cost and risk of the entire industry losing control — the very outcome Amodei says he is trying to prevent.
Part of a broader pattern of Chinese pushback
This is not the only recent flashpoint between Amodei and Chinese commentary. Anthropic’s own account of a prior controversy, where the company was accused of not de-deploying or fixing vulnerabilities in a model before US export controls took effect, drew a similarly pointed response from within the US policy world itself — former Trump administration official David Sacks argued that Anthropic’s handling of that episode was hard to square with its branding as “the AI safety company.”
Amodei has also faced pushback from a different direction within the US industry itself. When he previously argued against a ban on open-weight models in response to an open letter that left him isolated, it was reported that his underlying concerns — authoritarian governments building AI systems more powerful than the US and using that edge for military dominance — were largely the same concerns now driving the China provisions in “We Must Pace the Frontier.”
Where the debate goes next
Amodei has himself acknowledged that China is the central complication in his own proposal. In a separate interview covered by international outlets, he described what happens if adversarial nations decline to slow down alongside democratic ones as the “toughest dilemma” facing the entire framework — an admission that gives some ammunition to Global Times’ argument that the plan is less about universal safety and more about a race the US intends to win on its own terms before any slowdown talks with China begin.
For now, the essay has produced an unusual moment of alignment among US AI executives who normally spend much of their time criticizing each other, even as it has opened a new front in the broader US-China dispute over who gets to set the rules for how fast AI is allowed to advance.