How to lose the AI arms race: Tyler Cowen argues the US must accelerate, not nationalize
Mar 16, 2026
Key Points
- Tyler Cowen argues nationalizing AI development would hand the technological lead to competitors, since private companies built the strongest models and governments have historically failed to scale advanced software projects.
- Anthropic and the Department of Defense are in conflict over whether Claude should be banned from government work, with intelligence officials already using it for data interpretation and operational planning.
- Cowen's framework pairs tech acceleration with skepticism of government overreach, but the underlying tension remains unresolved: whether voters should have explicit voice in AI policy or whether speed of execution takes priority.
Summary
Tyler Cowen argues the US cannot afford to nationalize AI development or slow its progress. Either move would cost America the technological arms race.
The context is sharp. Leopold Aschenbrenner's essay "Situational Awareness" predicted AI companies would become functionally part of a government-led national security project, possibly nationalized. Senator Bernie Sanders called for a moratorium on AI data centers to slow progress. Anthropic and the Department of Defense are at loggerheads over whether Claude should be banned from government work, with intelligence officials already using it to interpret collected data and help plan operations.
The case against nationalization
Cowen grounds his argument in organizational capacity, not ideology. The Manhattan Project built the atomic bomb through top-down control, but the strongest AI models exist because the private sector built them. That remains true even in China, a far more statist country. The constraint is talent and salary. Top researchers command tens of millions in compensation, and governments have historically failed to scale advanced software projects. A shift to bureaucratic procurement would destroy the private companies driving progress and hand the lead to competitors.
Why the arms race is structural
AI weaponization is real. It enables cyberattacks, surveillance, bioweapon design, and potentially missile defense disruption. But competitive military technology races are structural to modern history. Space weapons, hypersonic missiles, and lasers would pose comparable threats regardless. AI simply accelerated what was already coming. The US currently holds the lead in AI, which plays to American strengths in creativity and private-sector innovation.
The prescription
Cowen proposes "tech accelerationism mixed with capitalism and then a prudent technocratic approach to military procurement" paired with skepticism of government overreach. This approach echoes 1960s-70s new left and libertarian anti-war sensibility. The goal is national security without becoming "the bad guys."
One tension remains unresolved. Framing AI policy as a battle between Anthropic CEO Dario Amodei and Pentagon official Pete Hegseth strips agency from voters. In a democracy, the American electorate should have explicit voice in AI policy the same way they assess whether a presidential candidate can be trusted with nuclear weapons. The question of who decides AI policy sits alongside the technical question of who can execute it fastest.