AI Arms Race Framing Threatens International Cooperation

The language of warfare has become ubiquitous in discussions of artificial intelligence development, particularly as competition intensifies between the United States and China. But according to Verity Harding, who spent four years briefing world leaders including Barack Obama and Emmanuel Macron on AI advances as head of global public policy at Google DeepMind, this framing may be fundamentally undermining the technology's safe development.

In a new essay anthology titled Reframing the AI Arms Race, Harding and contributors from politics and academia argue that casting AI as a weapon closes the door to international cooperation and forces smaller nations to align with one superpower or another, often against their own interests.

From collaboration to competition

Between 2016 and 2020, Harding's work at DeepMind centered on helping political leaders understand AI's capabilities and risks through a lens of international cooperation. That collaborative approach shifted dramatically, she notes, as rivalries emerged between individual labs like Anthropic and OpenAI, and between global superpowers.

The shift accelerated around ChatGPT's November 2022 launch, which coincided with heightened geopolitical tensions from the pandemic and Ukraine war. Two forces drove the change, according to Harding: genuine concern that the technology could be dangerous in the wrong hands, and an anti-regulation stance that weaponized China as a threat to argue against domestic oversight.

The Trump administration's nationalist AI rhetoric and recent export controls on frontier models exemplify how arms race thinking shapes policy, Harding told WIRED in a June interview.

The case for middle powers

Harding advocates for what she calls a "middle powers coalition" — countries like Canada, France, Japan, South Korea, India, and the UK working together rather than positioning themselves as chess pieces for larger players. Such a coalition could leverage India's scale and technology diffusion, the UK's talent and startup ecosystem, and Canada's critical minerals, among other complementary strengths.

The goal isn't to oppose national sovereignty in AI development, she emphasizes, but to recognize that competition and collaboration aren't mutually exclusive. Even the US and China cannot develop everything independently, creating strategic chokepoints around chips, minerals, and talent.

Why it matters

The arms race narrative isn't just rhetorical — it actively shapes how nations approach AI policy and international relations. By accepting the framing as inevitable, countries make it reality, potentially foreclosing paths to collaborative work on safety, security, and equitable distribution of benefits. For major labs, the arms race language concentrates power by positioning them as uniquely capable of managing an existential technology. The long-term risk, Harding warns, is excessive government centralization, less safe systems, and a world of vassal states forced to choose sides.

Harding also notes that major AI labs benefit from arms race rhetoric, which positions them as uniquely capable of managing powerful technology and accrues power to their organizations.

These details were first reported by WIRED in an interview with Harding discussing her essay anthology and the evolution of AI policy discourse.