This Former DeepMind Exec Thinks the AI Arms Race Could End in Disaster
Verity Harding tells WIRED that the US government’s nationalistic attitude toward AI is evidence that a worst-case scenario is taking shape.
Verity Harding tells WIRED that the US government’s nationalistic attitude toward AI is evidence that a worst-case scenario is taking shape. This rep
Read Full Story at Wired →Why This Matters
The warning from a former DeepMind executive underscores a critical inflection point in AI development: the collision between unchecked competition and existential risk. Nationalistic AI policies aren’t just reshaping geopolitics—they’re accelerating a global experiment with unknown safeguards, where the first-mover advantage could outpace our ability to mitigate catastrophic outcomes. This isn’t just about corporate or military advantage; it’s a systemic challenge to the stability of international relations in an era where code could outmaneuver diplomacy.
Background Context
The AI arms race traces its roots to the Cold War-era fascination with technological supremacy, but today’s iteration is fueled by private capital and data monopolies rather than state-run labs. Governments now treat AI as a strategic asset akin to nuclear deterrence, yet lack the institutional frameworks to govern its rapid deployment. Harding’s perspective is particularly sharp given DeepMind’s origins in cutting-edge research—its own evolution from a research lab to a Google subsidiary mirrors the blurred lines between innovation and industrial control that now define the field.
What Happens Next
If nationalistic AI policies harden into formalized blocs, we may see a bifurcation of standards—one set for domestic deployment and another for export—creating a patchwork of risks where misaligned systems could interact unpredictably. The most immediate flashpoint won’t be a headline-grabbing disaster but the erosion of shared oversight mechanisms, as agencies scramble to retroactively regulate tools that outpace their mandates. Watch for whether oversight bodies in the US, EU, and China can coordinate before the next breakthrough renders their frameworks obsolete.
Bigger Picture
This moment reflects a broader historical pattern where technological leaps outpace governance, from the industrial revolution to the atomic age, but with a twist: AI’s dual-use potential means the lag between invention and catastrophe is shrinking to months, not decades. The push for sovereignty in AI mirrors earlier resource wars, but the commodity this time is intangible—algorithmic advantage—and the battlefield is everywhere data flows. Without a new paradigm for multilateral control, the arms race isn’t just a metaphor; it’s the default operating system for the 21st century.
