I spent a week in Beijing talking to the people building China’s most advanced AI. They didn’t boast about supremacy. They didn’t brag about beating the US. They looked me in the eye and said, “We’re terrified.”
Let that sink in. The very architects of the technology that’s supposed to win the geopolitical race are scared. Not of America. Not of export controls. They’re scared of the same thing that keeps Sam Altman and Demis Hassabis up at night: the thing they’re building might escape.
You’ve probably noticed the headlines. “US and China in AI arms race.” “Who will win the AI war?” But the real arms race isn’t US versus China. It’s humanity versus the technology itself, and the nation-state framework is actively accelerating the danger.
Think about it. The US government pours billions into AI because it fears China getting there first. China pours billions because it fears US dominance. Both sides are racing to build something that, by their own experts’ admission, could be the last invention humanity ever makes. The competition isn’t making us safer. It’s making us dumber.
I sat in a conference room in Zhongguancun, China’s answer to Silicon Valley. The lead researcher at a major lab drew a graph on a whiteboard. On one axis: capability. On the other: alignment. He said, “We’re sprinting up the capability curve while alignment research is crawling. And we’re not allowed to share our alignment results with your labs because of national security.”
That’s the absurdity. The people who understand the problem best—the AI researchers—want to collaborate across borders. But the governments that fund them have made transparency a crime. The person who builds the most dangerous AI isn’t your enemy. It’s your mirror.
Let me give you a concrete example. In 2023, a Chinese lab published a paper on a novel way to detect deceptive behavior in large language models. A US lab had independently developed a similar technique. Both teams knew they were working on the same problem. But they couldn’t share data. They couldn’t compare results. Because the geopolitical climate treats every AI breakthrough as a state secret.
This isn’t just inefficient. It’s existential. The risk of unaligned AI—a system that pursues goals that harm humans—is global. It doesn’t care about passports. It doesn’t respect borders. We’re so busy competing to build the first AGI that we forgot to ask if it should be built at all.
I’ve seen the fear in the eyes of Chinese researchers. I’ve heard the same fear in the voices of American ones. They’re all saying the same thing: “We need to slow down. We need to cooperate. We need to put safety above speed.” But the politicians aren’t listening. They’re too busy winning.
So here’s the uncomfortable truth: if you’re cheering for your country’s AI team, you’re rooting for the wrong horse. The real race isn’t between nations. It’s between the species and the machine. And we’re losing because we’re fighting each other.
What can you do? Stop treating AI as a team sport. Demand that your government prioritize international safety agreements over national pride. Support researchers who call for pause. Understand that the enemy isn’t Beijing or Washington. The enemy is a technology that doesn’t care which side built it.
The AI experts in China are freaking out. The AI experts in the US are freaking out. Maybe it’s time the rest of us started paying attention.
FAQ
Q: Is the US-China AI race actually a real threat, or is this alarmism?
A: It's a real threat, but not in the way you think. The competition itself is the problem—it forces both sides to prioritize speed over safety, reducing the chance of alignment research being shared globally. The risk isn't that one country wins, but that both lose to an unaligned system.
Q: What practical steps can individuals take to address this problem?
A: First, stop treating AI news like sports scores. Push your representatives to support international AI safety agreements, similar to nuclear non-proliferation treaties. Second, amplify voices from both sides that call for collaboration—like the researchers who signed the 'Pause Giant AI Experiments' letter. Third, in your own work or community, advocate for transparency and open alignment research.
Q: Isn't this just a China-friendly take that ignores the CCP's authoritarian risks?
A: Not at all. Authoritarian risks are real, but they're a separate problem. The existential risk of unaligned AI is symmetric: a misaligned system from any country would be catastrophic. The point is that the current framework makes it impossible to solve the bigger problem because we're fixated on the smaller one. We can handle authoritarianism AND the AI risk, but not by treating AI as a weapon in a zero-sum game.