An AI Slowdown Would Hand China the Future

All that anyone can talk about is AI, and for good reason. Since January 2025, Anthropic and OpenAI have collectively grown their ARR to over $70–$90 billion, according to disclosed and third-party estimates. This technology is everywhere, whether you’re an accountant working for a firm servicing small businesses in the Midwest or a senior software developer working for a hyperscaler like Amazon or Meta. Despite numerous examples of the good this technology can provide—like identifying cancer early or solving previously intractable problems in math and physics—it is unequivocally despised by the American populace.

It is now all but confirmed that there’s been a concerted effort by numerous foreign groups and organizations to instill hatred and distrust among Americans regarding AI, for reasons that should be obvious: It’s a revolutionary technology, and America would be worse off without it. Regardless of where it originated, the idea has taken hold—but this mindset has existed for some time now, and the distinctions between AI safety activism and vocally anti-AI populism are now hard to distinguish in principle.

The risks of AI should not be understated, and research done by those in the field of AI safety and existential risk should absolutely not be discounted. But what good does it do us as a country to pontificate on the negative externalities of a technology over the long term while foreign adversaries like China advance at a breathless pace (as it has just done with the release of a model like Kimi K3)?

Kimi has made history as the first non-American AI model to surpass all publicly available frontier models in significant capabilities and benchmarking results, breaking the assumption that compute constraints would lead to an American victory in the AI race or stall Chinese progress.

Prominent AI researchers tout the work of a small research group—the AI Futures Project—and, more recently, proclaim the gospel of its latest essay and thought experiment, AI 2040, a legitimately novel look at what the short and near term in AI development could look like. As comprehensive and rigorous as it is in its approach, there should simply be no attention given to the feasibility of an AI slowdown in the short term, given the stakes at hand, which range from net-new job creation and GDP growth to matters as existentially important as transformative AGI and dramatic economic growth within a short timespan.

The AI 2040 experiment begins with a mutual agreement between the U.S. and China—the two de facto leaders in AI globally—to institute a more cautious and transparent approach to AI development, culminating in things like a halt to AI training, compute declarations, and total research transparency, all in the service of our planet and its global citizenry after we purportedly assume AI progress is an existential threat to global prosperity.

The main idea is that AI development will continue to accelerate at such a ferocious pace that there is nothing left for us to do but surrender decision-making to policymakers, under the generous assumption that government officials and governmental bodies will understand such a transformative technology better than the individuals developing it. Such a proposition would make sense if there were any evidence to suggest that AI was actively undermining the ability of humans to prosper and succeed in society. Yet evidence shows that AI has been nothing but beneficial to its early adopters and embracers across economic statuses and levels of industry.

Additionally, AI 2040 assumes that regulators will act in good faith, only determining that an AI slowdown is an optimal solution after “private AI companies iteratively solve the alignment problem,” as if the government were watching this industry with bated breath to determine the most game-theoretically optimal move before making a decision. In reality, various sectors of the U.S. government, such as the administration of New York Governor Kathy Hochul, have signed harmful executive orders outright banning the development of data centers, citing a shaky bogeyman involving widespread public opposition and rising energy costs.

Much of the AI 2040 essay depends on an initial assumption that China will come to terms with the U.S. on a mutual proposal to stop AI development. While ambitious and coming from a place of good faith, the writers seem to be unaware of the intricacies of global politics and our sensitive relationship with China, exacerbated by Donald Trump’s July 16 State of the Union address, which illuminated Chinese interference in American elections and definitively labeled China an enemy before 400 million Americans.

In President Xi Jinping’s speech at the World AI Conference, he emphasized the importance of AI, a commitment to open source, and the importance of not overstretching national security in the discussion of AI.

Any call to pause AI development—an industry primarily dominated and dictated by American companies like OpenAI and Anthropic—should come across as a declaration of defeat in the now-imminent global AI race. A country like China, notorious for its “Five-Year Plans” for technological development and dominance, should be considered nothing less than an enemy and far from a willing partner in kneecapping technological progress in the name of safety from a technology that has yet to produce or suggest harm against humans.

In an ideal world, it would make sense to gather around and declare such a pause on further AI development, but we will never find ourselves occupying such a world. Global society and geopolitics are far more complex than ever before in human history, and there’s little evidence to suggest this acceleration will cease to continue.

A technology as existentially important to human progress as AI shouldn’t be treated as a weapon, but as a tool for continued American prosperity far into the future. Because if we don’t secure this future, someone else will gladly pick up where we fatefully chose to leave off.