Former DeepMind exec warns AI arms race framing risks global disaster

Verity Harding, who served as Google DeepMind's head of global public policy between 2016 and 2020, has witnessed the transformation of AI discourse from collaborative science to a zero-sum geopolitical contest. In a new essay anthology she curated, *Reframing the AI Arms Race*, Harding and contributors like historian Lawrence Freedman and Japanese politician Taro Kono argue that the dominant metaphor of an "arms race" is not just inaccurate but actively harmful. During her tenure, Harding briefed world leaders including Barack Obama and Emmanuel Macron on AI's potential and risks, emphasizing international cooperation. However, she observed a shift fueled by two forces: genuine fear that AI could become a weapon in hostile hands, and an anti-regulation narrative that paints China as a bogeyman to justify deregulation.

The consequences of this framing, Harding warns, are already visible in policy and rhetoric. She points to the Trump administration's nationalist AI stance and its push for export controls on U.S. models as symptoms of an arms-race mentality that closes the door to multilateral governance. Smaller nations that import AI technology are forced to align with either the U.S. or China, often against their own interests. Harding argues that the metaphor restricts creative thinking about safety, distribution of benefits, and democratic oversight. She calls for a return to the internationalist approach that defined early AI policy work, warning that treating AI as a weapon risks accelerating a race to the bottom where no one can guarantee safety.