The most consequential technology competition of the century is being fought by two nations that cannot even agree on what they're afraid of.
While American discourse around artificial intelligence has become saturated with existential dread—rogue superintelligences, alignment failures, the specter of human obsolescence—China's approach to AI risk operates on an entirely different frequency. Beijing's concerns are practical, immediate, and deeply political: information control, social stability, economic disruption. The apocalyptic scenarios that animate Silicon Valley boardrooms and Washington hearings barely register in Zhongnanhai.
The anxiety gap
This divergence is not merely philosophical. It produces radically different regulatory environments, investment priorities, and development timelines. American AI companies increasingly operate under the shadow of potential liability for harms their models might theoretically cause decades hence. Chinese firms face intense scrutiny over what their models say about Tiananmen Square or Xinjiang today.
The result is a kind of regulatory arbitrage. While OpenAI, Anthropic, and Google navigate an increasingly cautious domestic landscape—voluntary commitments, red-teaming requirements, the looming possibility of federal legislation—Chinese competitors like Baidu, Alibaba, and a constellation of state-backed startups operate with different constraints entirely. They worry less about hypothetical superintelligence and more about ensuring their chatbots never contradict Xi Jinping Thought.
Different fears, different futures
The American preoccupation with existential risk reflects a particular intellectual tradition: the effective altruism movement, the rationalist community, the long-termist philosophers who have captured disproportionate influence over how the tech industry thinks about AI. These frameworks emphasize low-probability, high-magnitude catastrophes—scenarios where a misaligned AI might, in the worst case, end human civilization.
China's Communist Party has little patience for such abstractions. Its concerns are rooted in a century of political instability and a governing philosophy that prioritizes regime continuity above all else. An AI that can organize protests is more frightening than one that might theoretically become conscious. A language model that undermines party narratives poses a clearer danger than one that might someday exceed human intelligence.
This isn't to say Chinese researchers ignore technical safety. They don't. But the framing differs fundamentally. Safety means controllability. Alignment means alignment with state objectives.
The strategic calculus
For American policymakers, this asymmetry presents an uncomfortable dilemma. Aggressive domestic regulation might address genuine risks but could also cede ground to competitors operating under fewer constraints. The precautionary principle, applied unilaterally, becomes a form of strategic disarmament.
Yet the alternative—a race to the bottom on safety standards—carries its own dangers. The nations developing the most powerful AI systems are also the ones that will shape the norms governing their use. If China's model of AI governance prevails globally, the technology's future will be shaped by priorities that place regime stability above individual rights or democratic accountability.
Our take
The uncomfortable truth is that both nations are probably worried about the wrong things. America's fixation on speculative superintelligence risks distracts from AI's more immediate harms: discrimination, surveillance, labor displacement, the erosion of shared reality. China's obsession with ideological control blinds it to the brittleness of systems built to serve propaganda rather than truth. The AI race is being run by competitors who have each convinced themselves they understand the finish line. Neither does.
