The AI arms race is a lie we’re telling ourselves. And it is expensive.
Verity Harding knows this better than most. Between 2016and2020, she wasn’t just coding algorithms. She was briefing world leaders. Barack Obama listened to her. So did Emmanuel Macron. As the head of public policy at Google DeepMind Harding spent years trying to keep things civil.
Back then? International cooperation felt possible.
“AI research was rooted in international cooperation.”
Then something shifted. The collaborative vibe vanished. It was replaced by rivalry. Anthropic versus OpenAI. The US versus China. Suddenly everyone was talking about a war. An arms race. The phrase stuck like burrs in wool.
In Reframing the AI Arms Race, Harding and others like historian Lawrence Freedman argue that language shapes policy. Words matter. Calling AI a “lethal weapon” might sound dramatic, but it kills diplomacy. It shuts down the cooperation needed to keep the tech safe and distributed fairly.
For smaller nations the stakes are higher. If they buy into the arms race narrative, they have to pick a side. They become pawns. Either the US or China wins. They rarely get a vote.
The Sexy Lie
Harding tells WIRED the war metaphor is popular for one simple reason: it feels clarifying. It’s sexy. Simple narratives are comfortable. Dig deeper though, and it restricts your thinking entirely.
Why the shift from “exciting science” to “civilizational battle”?
Two things happened.
First. People got scared. There’s a genuine fear that AI in the wrong hands is catastrophic. Democracies, the thinking went, need to hold the keys. Control must stay here. Not there.
Second. Anti-regulation voices found a handy villain. Pointing at China as the “bogeyman” made deregulation seem patriotic.
If you regulate, China wins. That’s the pitch.
Then ChatGPT arrived. Late 2022.
The timing couldn’t have been worse for sanity. The world was already reeling. Pandemics made borders feel fragile yet urgent. The war in Ukraine turned abstract geopolitical theory into mud and blood. Suddenly, AI weaponry wasn’t a sci-fi concept. It was real.
The narrative solidified instantly. AI became the new nuclear option. The new Cold War. History rhymed in the worst way.
Isolationism Wins
Who controls technology when superpowers fight? The answer is rarely clear, but the chaos favors the loud ones.
Tech shapes society, sure. But society shapes tech just as hard. Right now, the tense political climate in the US is dictating how AI develops. Isolationism is driving policy.
Harding argues that turning inward is a bad strategy. Sovereign capacity in the UK and Europe is vital. Yes. But total isolationism obscures reality.
Not even the superpowers can build everything themselves. The US needs chips. China needs critical minerals. Everyone needs scientists. The supply chain is a series of strategic chokepoints.
“You can’t have our chips.”
“Well, you can’t have ours either.”
It’s unrealistic to assume any country can maintain a completely sovereign AI stack. The dependencies are too tangled.
The Middle Powers
The Trump administration leaned hard into this nationalist rhetoric. An executive order drenched in American-first ideology forced Anthropic to pull its latest model. It sent a shockwave through the industry.
European powers should be worried. They rely heavily on US tech.
But cooperation and competition aren’t enemies. Harding suggests a middle ground. A coalition of middle powers.
Think about it. Canada has the minerals. France and the UK have the talent and the ecosystems. Japan and South Korea have the engineering depth. India brings massive scale.
Together, they have leverage. Together, they have scale.
The point is to not allow the arms rate framing to convince you that the game is a binary race.
When small nations believe they are just chess pieces in a binary struggle, they make it true. They become lesser players. Accepting that premise surrenders agency.
Who Benefits?
Money rushes into fear. The speed of capital injection into AI has been frenetic. That velocity pushed the narrative. But cash isn’t the only corruptor.
The big labs? They are complicit.
Framing AI as an exclusive, high-stakes weapon gives power to those who hold the tools. It implies the tech is too dangerous for anyone but the giants. Too complex for regulation. Only they know how to fix it. Only they can lead.
It’s a self-serving story. A way to entrench control under the guise of security.
The race framing convinces us there are only two runners. The reality is much more complex. We just haven’t looked at the finish line clearly in years. Maybe it’s time we did. Or maybe we just keep running.
