AI Photo: VCG

Not long ago, US artificial intelligence (AI) executives were championing the technology's power. Over the past several days, however, they have shifted gears, calling for a slowdown on safety grounds.

What drove this shift in narrative? What are these leading US AI firms actually doing to address legitimate safety concerns about the technology? Would a so-called coordinated slowdown in AI research actually work? And what does this narrative have to do with the intensifying AI competition between China and the US?

To answer these questions, the Global Times examined the arguments made by US AI executives, combed through extensive reports on their firms' latest developments, and conducted interviews with veteran Chinese tech analysts. The investigation reveals an ulterior motive - maintaining market dominance - rather than altruistic concerns about safety. Worse still, analysts noted, such narratives advanced in the name of safety actually pose real risks of confrontation and vicious competition.

Vicious competition

The latest buzz began on September 12, when Anthropic CEO Dario Amodei published an essay arguing that leading AI labs "must slow the pace at which we improve the capabilities of AI models." Within hours, Elon Musk, head of xAI, replied on X: "Dario is right." OpenAI CEO Sam Altman also claimed that OpenAI would adopt similar steps. The rare public alignment among three fierce rivals in the US quickly made global headlines and sparked a debate about the progress of AI development.

While legitimate safety concerns around AI require global discussion, the latest narrative from these US AI executives does not seem to be a genuine attempt at it. After all, the same companies are accelerating on every front that matters: computing power, new models, IPO preparations, and rule-making. Analysts see the collective call for a slowdown as a strategic move for commercial interests rather than pure altruism.

The collective call is "a sophisticated move by the front-runners to convert short-term tech leads into lasting institutional barriers [for other firms]," Zhang Xiaorong, director of the Beijing-based Cutting-Edge Technology Research Institute, told the Global Times. The US AI executives, he said, are trying to use the safety slowdown narrative "to embed their advantage into the industry's rules - completing the shift from racers to rule-setters." 

If a company truly believed the frontier innovation should slow down, its behavior would match. Anthropic's does not, according to various media reports about the firm's latest moves.

Reuters reported that Anthropic is preparing what could be the largest IPO in history, seeking to raise $100 billion at a valuation of around $2 trillion. Amodei's essay appeared the day after those plans were disclosed.

"The slowdown narrative fits Anthropic's IPO timing. It could reframe any possible softening in revenue growth as 'responsible leadership' on safety," Zhang said, pointing to huge sums spent on computing power investments.

In a short period, Anthropic has locked in massive computing resources for the next decade, including a $35 billion deal with Nvidia-backed Lambda, a $45 billion agreement with Nscale, monthly payments of about $1.25 billion to SpaceX, and commitments of more than $100 billion to Amazon Web Services over 10 years.

The Wall Street Journal also noted the contradiction: while Amodei urges the industry to slow advanced development, Anthropic has not paused or slowed its own work.

This is also true for OpenAI, which endorsed the slowdown. The US AI firm released GPT-6 Astra in early September, and described it as "the world's most intelligent and aligned model." 

Many in the US also saw through this. "Stop pretending the motive for slowdown is purely altruistic," former White House AI official David Sacks wrote on social media, calling the proposal "a regulatory capture."

Media reports also noted that the leading AI firms propose regulation in order to seize rule-making power, avoid stricter government oversight, and freeze the existing hierarchy. The pattern recalls Nobel laureate George Stigler's theory of regulatory capture, in which industries seek rules that protect their own interests.

"[The calls for] higher safety testing, third-party audits, and compliance requirements would raise barriers further. Established industry leaders can absorb those costs; later entrants and smaller players would struggle," Liu Dian, an associate researcher at Fudan University's China Institute, told the Global Times. "The practical effect would be to slow competitors more than the leaders themselves."

"At its core, it is a fight for the power to define the rules of the global AI industry," Wang Peng, an associate research fellow at the Beijing Academy of Social Sciences, told the Global Times.

Promoting confrontation

Beyond the vicious competition, there is also a clear risk of promoting confrontation over technology.

In his essay, Amodei wrote that a Chinese lead in AI "would pose grave danger." He explicitly supported continued advanced-chip export controls on China and stronger measures against model distillation, arguing that "democratic nations" must maintain their lead. He further proposed that the US and other countries should try to bring China under similar restrictions on the pace of AI development.

Chinese analysts see this framing as far from genuine safety concern. Instead, they view it as an attempt to lock in America's commercial and technological advantage by turning safety into a geopolitical tool.

"These demands are framed as about AI safety, but essentially expose the desire to preserve US AI advantage, consolidate technological hegemony and clear competitive obstacles for global expansion," Liang Huaixin, a researcher at the University of International Business and Economics' National Security and Governance Institute, told the Global Times.

China's Foreign Ministry spokesperson Guo Jiakun on September 14 stated that AI is a consequential technology for the wellbeing of all humanity, and all parties should jointly promote the open and inclusive development of AI for good and for all. 

"Fear-mongering, confrontation and vicious competition will only hamper efforts toward sound global AI governance, which serves no one's interest," Guo said, in responding to a question about Amodei's argument that a Chinese advantage in AI would pose US national security risks.

While US companies attempt to set the pace by erecting barriers for others, China's open-source approach is rewriting the competitive landscape with a different logic. 

China's open approach

"China, as the world's largest and most open supplier in the global AI ecosystem, has long ceased to be a variable that can be 'excluded.' Chinese open-weight models allow users to download and customize them with their own data, lowering barriers and challenging the technological and price monopoly long held by a small group of US firms," said Wang.

Despite the US' restrictions, Chinese open-source models continue to gain ground. Hugging Face's Spring 2026 report showed Chinese models accounting for 41 percent of downloads on the platform, surpassing the US share. Alibaba's Qwen series has generated more than 113,000 derivative models. On the model-routing platform OpenRouter, the top models by clicks have frequently come from Chinese labs, reflecting strong global demand for open-source and lower-cost alternatives.

US startups and even some larger firms have already adopted Chinese open models for cost and performance reasons. Notably, when reports of possible US restrictions surfaced, nearly 200 Silicon Valley companies publicly opposed the idea.

Attempts to erect barriers in the name of safety cannot stop China's open-source development model, nor can they prevent Chinese AI from winning growing recognition and adoption around the world, Chinese analysts noted.

Meanwhile, China is also sharing the AI dividends with all over the world, showcasing its advocacy of "developing AI for good and for all." 

"AI should not be owned by major countries, still less dominated by contest and rivalry. China believes in a people-centered approach to AI and advocates openness and inclusiveness to ensure AI is a force for good and for all," Mao Ning, a spokesperson for the Foreign Ministry, said in June.