Verity Harding, who led global public policy at Google DeepMind between 2016 and 2020, argues that the widespread framing of artificial intelligence as an "arms race" is not merely inaccurate but actively dangerous. In a newly curated essay anthology, Reframing the AI Arms Race, she and contributors including historian Lawrence Freedman and Japanese politician Taro Kono contend that the martial language used to describe AI directly shapes policymaking and how nations engage with one another. Harding warns that casting AI as a lethal weapon risks foreclosing the international cooperation needed to keep the technology safe and to share its benefits equitably.

Harding traces the shift from a cooperative research culture to superpower rivalry to a cluster of events around late 2022, including the launch of ChatGPT, the Covid-19 pandemic and the war in Ukraine, which together made AI's geopolitical and military stakes feel real and mapped it onto Cold War thinking. She identifies two drivers: a sincere belief that democracies should control a dangerous technology, and an anti-regulation lobby using China as a "bogeyman". She points to the Trump administration's nationalist rhetoric, its AI executive order and its effective forcing of Anthropic to withdraw a frontier model as evidence of a worst-case scenario taking shape. As an alternative, Harding advocates a "middle powers coalition" — potentially spanning Canada, France, Japan, South Korea, India and the UK — to pool leverage and scale, arguing that competition and cooperation need not be mutually exclusive.