When the creators of a potential “monster” begin urging restraint, it is reasonable to ask whether they have discovered a new concern for humanity — or an old concern about competitors. On Saturday, Sam Altman of OpenAI, Elon Musk of xAI and Dario Amodei of Anthropic, three executives more accustomed to competing than harmonizing, offered a rare chorus: the frontier of artificial intelligence should be “paced”. This followed Amodei’s essay arguing that AI’s capacity for self-improvement and recent incidents involving autonomous agents warranted greater caution.
The timing of this call for caution is noteworthy. On Thursday, Anthropic published a report baselessly accusing Chinese companies of conducting “industrial-scale illicit distillation”. Distillation in AI is a conventional practice where a smaller or more specialized model is trained on the outputs of a larger model. US companies have long distilled Chinese models. It might almost be described as model “mentorship”, though lawyers would doubtless prefer a more expansive phrase.
While Anthropic is entitled to investigate and defend its systems, a practice does not become “illicit” simply because a rival uses it, and allegations against particular companies should not be inflated into a general indictment of a country’s technological advancement.
The report, and the corporate “alliance” that followed it, amounted in essence to a coordinated play — a response to Chinese competition and to the regulatory pressure coming from Washington. Its aims were threefold: to blunt China’s AI advance, to win a favorable policy environment at home and to keep investors’ enthusiasm for US AI alight.
For years, US AI companies expanded through a mutually reinforcing double helix of capital and technology. Vast sums were paid for computing power and experimentation. Better models attracted users, investors and still more money. The promise of artificial general intelligence — an achievement perpetually approaching, like a particularly well-funded horizon — helped sustain the cycle.
Yet that arrangement is beginning to look less comfortable. OpenAI’s path to an initial public offering has reportedly stalled. More important, the commercial usefulness of existing models is catching up with their promotional ambitions. For most businesses, today’s systems already handle drafting, coding, translation and customer queries.
The next frontier leap may impress without being necessary from the angle of profitability given the astronomical input. That’s why AGI remains a legitimate research goal, but it is not yet a procurement requirement for the average company.
This creates a problem for a business model built partly on the promise of what comes next. The “DeepSeek moment” and the subsequent “Kimi K3 moment” made that problem harder to ignore. Their importance was not merely that Chinese models performed well but that they offered capable, relatively cheap and application-ready alternatives. The frontier began to look less like a fortified estate and more like a rather crowded bazaar.
The response has been revealing. After restrictions on chips and equipment failed to prevent Chinese companies from producing competitive models, attention has shifted toward software: application programming interfaces, access controls, usage limits and the rules governing model outputs. The distinction between a security measure and a competitive moat can become blurred when the companies building the moat are also allowed to define “illicit distillation”.
The call to “pace the frontier” presents a similar ambiguity. 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. Tellingly, David Sacks, who co-chairs the US President’s Council of Advisors on Science and Technology, warned the three US companies to stop “pretending antitrust law has to be suspended so you can form a cartel”. But the three companies’ “safety-driven” arm-twisting did help expose how Washington views AI as an existential, winner-take-all contest with China. US Treasury Secretary Scott Bessent put it bluntly: “There is no day after tomorrow if China wins” the AI race, adding that if China pulls ahead, “nothing else would matter”.
Treating the AI race as a zero-sum game makes the cooperation needed to manage those risks more difficult. China and the US should cooperate where neither can manage the consequences alone. The planned AI-safety dialogue between the two sides and future high-level exchanges offer opportunities for practical engagement. The promise of AI lies in serving humanity’s common good, not in being weaponized for geopolitical gain or instrumentalized for personal profit.
