The current cultural conversation surrounding artificial intelligence is dominated by a single, pervasive metaphor: the “arms race.” We see it in the media, hear it in congressional hearings, and feel it in the boardroom anxieties of top tech executives. Whether it is depicted as a high-stakes standoff between individual labs like OpenAI and Anthropic, or as a binary Cold War-style struggle between the United States and China, this framing has become the default way to understand the evolution of intelligence. However, Verity Harding, a former head of global public policy at Google DeepMind who spent years advising world leaders like Barack Obama and Emmanuel Macron, argues that this narrative is not merely a descriptive shorthand—it is a dangerous catalyst for the very worst-case scenarios we are trying to avoid.
During her time at the forefront of AI policy, Harding witnessed a profound shift in tone. In the mid-2010s, the development of AI was largely characterized by a spirit of international cooperation and shared scientific inquiry. There was a genuine, collective interest in mapping out the ethical traps and potential safety risks of a technology that belonged to humanity at large. But as the landscape evolved, the discourse soured. A “civilizational battle” narrative took root, fueled by a mixture of genuine fear regarding the technology’s potential for harm and a strategic, often cynical, narrative pushed by those wishing to avoid regulation. By casting AI as a lethal weapon, policymakers have inadvertently limited the horizon of possibility, making the collaborative governance of AI seem not just difficult, but treasonous.
This “arms race” framing carries severe consequences, particularly for the global distribution of power. When AI is viewed strictly through a military lens, it forces smaller nations into a precarious position: they must either align themselves with one superpower or risk being left behind in a technological vacuum. This forces countries to act against their own long-term interests, tethering their infrastructure to the geopolitical whims of Beijing or Washington rather than to their own national priorities. Harding points to the rise of nationalist rhetoric and aggressive export controls as direct, unfortunate symptoms of this framework. By treating advanced computing power as a resource to be hoarded and weaponized rather than a scientific achievement to be shared, we are effectively breaking the global consensus necessary to manage the technology safely.
The origin of this obsession with state-versus-state conflict can be traced back to the human tendency to reach for familiar patterns. In a world reeling from a global pandemic, border-tightening, and the kinetic reality of the war in Ukraine, the “Cold War” template was easy to overlay onto AI. The launch of ChatGPT in late 2022 acted as a force multiplier for this anxiety, turning theoretical research into a “Sputnik moment” in the public eye. People began to view AI not as a tool for progress, but as a nuclear-grade asset. As Harding notes, this narrative is admittedly “sexy”—it provides a sense of urgency and clarity—but it is ultimately restrictive. It shuts the door on the international coordination that would allow us to actually govern AI, instead incentivizing a race to the bottom where safety is sacrificed in the name of speed and competitive advantage.
Harding’s new anthology, Reframing the AI Arms Race, features contributions from esteemed historians and international policymakers who reject this fatalistic view. By challenging the linguistic cage of the “arms race,” they hope to broaden the terms of engagement. They argue that if we continue to define AI as a weapon, we will ensure it becomes one. Instead of viewing the potential of AI as a zero-sum game where one nation’s gain is another’s loss, these thinkers advocate for institutionalized cooperation. They suggest that the only way to ensure the benefits of AI are evenly distributed—and that the risks are kept in check—is to move away from the bogeyman of geopolitical rivalry and toward a model of collaborative oversight that includes nations currently being treated as bystanders.
Ultimately, the challenge of this era is to see through the “sexy” narrative of conflict and reclaim the original promise of AI research. We stand at a junction where our choice of language directly dictates our future. If we insist on calling it a race, we will run blindly toward a cliff; if we can reframe it as a shared global challenge, we might actually build the guardrails necessary to thrive. Harding’s call to action is a plea to look beyond the immediate political gains of alarmism. By refusing to accept the “arms race” as an inevitable reality, we open up the possibility of a world where AI serves as a bridge between nations rather than a wall, ensuring that the technology of the future is defined by our collective wisdom rather than our mutual fear.
