Former DeepMind Exec Warns AI Arms Race Could End in Disaster

The language we use to describe artificial intelligence may be more dangerous than the technology itself. When nations and tech giants frame AI development as an arms race, they risk locking the world into a cycle of confrontation that undermines safety, cooperation, and equitable access to progress.

That is the central argument from Verity Harding, a former head of global public policy at Google DeepMind. From 2016 to 2020, Harding briefed world leaders including Barack Obama and Emmanuel Macron on AI’s trajectory. During that period, she recalls, AI research “was rooted in international cooperation.” But the narrative shifted. Competition between labs like Anthropic and OpenAI and the escalating rivalry between the United States and China transformed the conversation. The AI arms race became the dominant metaphor.

In a new essay collection she edited, titled Reframing the AI Arms Race, Harding assembles voices from global politics and academia,among them historian Lawrence Freedman and Japanese politician Taro Kono,to argue that the way we talk about AI directly shapes policy and international relations.

Treating AI like a weapon, Harding warns, closes the door to the kind of global collaboration needed to ensure the technology remains safe and its benefits are broadly shared. For smaller nations that import AI systems, accepting the arms race framing means choosing sides between superpowers, often against their own economic and strategic interests.

Harding points to the Trump administration’s nationalist AI rhetoric and its push for export controls on American models as symptoms of this dangerous framing. To her, these moves are evidence that the worst-case scenario is already taking shape.

WIRED spoke with Harding in early June about where the arms race idea came from, how it is reshaping geopolitics, and what smaller countries can do to secure a seat at the table. The conversation has been condensed and edited for clarity.

WIRED: Why do people gravitate toward war metaphors when discussing AI?

VERITY HARDING: It’s a seductive framing. It feels clarifying, but if you look closely, it actually limits your thinking.

When I worked at DeepMind, our job was to help political leaders understand what AI could do. That work was grounded in the idea that the technology was exciting but also raised concerns that were best addressed collaboratively, across borders. Over time, though, I noticed the conversation shifting toward a civilizational battle: the West versus China.

What drove that change?

Two things. First, a genuine belief that AI was dangerous,or could be in the wrong hands,and that democracies needed to control it. Second, an anti-regulation camp that found it useful to paint China as a bogeyman. Their argument was: “If you regulate us, you let China win.”

Was there a specific turning point?

ChatGPT launched in November 2022, and it made a huge number of people suddenly pay attention to AI. But other events were unfolding at the same time. The pandemic had just made the world feel both borderless and fiercely bordered again. Then came the war in Ukraine, which made discussions about AI and geopolitics,especially weaponry,feel suddenly urgent and real.

Very quickly, it became accepted wisdom that AI is the new arms race. People mapped it onto the last major arms race they remembered: the Cold War. They talked about AI as if it were a nuclear weapon.