AI Nationalism Prioritizes Escalation Over Safety, Warns Expert

The public warning from former DeepMind executive Verity Harding against a rising tide of AI nationalism reframes the industry's central conflict away from corporate rivalries and toward state-level competition. This shift, exemplified by the US government’s increasingly nationalistic rhetoric, matters immensely as it provides a geopolitical justification for prioritizing capability escalation over collaborative safety research. As nations like the US and China enact policies to achieve AI dominance, such as sweeping chip export controls, Harding’s insider critique lends critical weight to fears that a dangerous, zero-sum mindset is displacing the open, global collaboration that defined AI's last decade of progress. The mechanics of this AI "arms race" create a perilous feedback loop where perceived threats from rival nations justify accelerated, risk-prone development and massive state subsidies. The primary winners are not consumers or even the AI labs themselves, but defense contractors and critical hardware suppliers like NVIDIA, who are positioned to capture enormous budgets allocated under the banner of national security. This dynamic forces a strategic recalculation for all players, fundamentally altering the risk-reward equation. The losers are open-source initiatives and international research partnerships, which are increasingly viewed with suspicion, threatening to fragment the global AI ecosystem into siloed, competing blocs. Looking forward, this trajectory points toward a balkanized AI world. In the next 12-18 months, expect to see more explicit "friend-shoring" of AI development, with Western governments discouraging collaboration with researchers in China, and vice-versa. The critical variable will be whether a major AI-driven geopolitical incident—an "AI-Cuban Missile Crisis"—can be avoided before international safety treaties gain traction. Harding’s warning suggests the current path, driven by nationalistic fervor, makes such a crisis more probable, arguing that the greatest existential risk is not rogue AGI, but state actors deploying powerful, unconstrained AI for geopolitical gain.