Verity Harding: Treating AI as an Arms Race Threatens Global Safety

The prevailing narrative that artificial intelligence development has become a weaponized competition between the United States and China may be doing more harm than good. That's the core argument from Verity Harding, who spent years as head of global public policy at Google DeepMind advising world leaders on AI's trajectory and implications.

In a new essay collection titled Reframing the AI Arms Race, Harding and contributors including historian Lawrence Freedman contend that the language we use to describe technological advancement directly shapes how nations negotiate with each other and craft policy.

How the Narrative Shifted#

When Harding was advising politicians between 2016 and 2020, AI policy discussions centered on shared challenges and opportunities for coordinated international effort. That tone shifted dramatically around 2022, when public generative AI systems like ChatGPT captured mass attention—a moment that coincided with renewed geopolitical tensions, pandemic fallout, and the Russia-Ukraine conflict. The media and policymakers swiftly adopted a Cold War metaphor, comparing AI supremacy to nuclear weapons development and framing the competition as existential.

Harding identifies two forces driving this reframing. One reflects genuine concern that advanced AI could pose risks if developed without robust safeguards—leading some to argue that democratic nations must maintain technological leadership. The other exploits these fears instrumentally: companies and governments opposed to regulation point to China as a threat, implying that oversight would cede advantage to a rival.

The Dangers of Seeing AI as Inevitable War#

Casting development as an arms race creates self-fulfilling prophecies. Nations and corporations internalize the competitive framing and abandon cooperation, even on issues where collaboration would strengthen safety and responsibility. The language also consolidates power: major AI labs benefit from the narrative that only they possess the expertise to navigate such weighty competition.

Smaller countries and middle powers face particular risk. If they accept the binary framing—picking a superpower patron or facing obsolescence—they forfeit meaningful agency in shaping how AI affects their citizens. They become passive consumers or strategic pawns rather than participants.

An Internationalist Alternative#

Harding advocates for a coalition of middle powers—potentially including France, Japan, South Korea, India, Canada, and the UK—that collectively possess leverage without the illusion of autarky. India brings distribution scale, the UK has research talent and startup ecosystems, Canada controls critical minerals. By pooling resources and negotiating as a bloc, such countries can resist the binary-race framing and insist on collaborative approaches to security, food systems, and health.