America’s Superintelligence Dilemma: How to Avoid an AI Catastrophe

Foreign Affairs | Hal Brands

In 2026, accelerated developments in artificial superintelligence (ASI) and autonomous military systems have forced the United States to confront severe geopolitical and existential risks. Emerging capabilities in superhuman cyberwarfare, self-improving models, and loss-of-control episodes threaten to upend the global balance of power while raising the potential for catastrophic consequences.

This technological acceleration presents Washington with three distinct strategic pathways: pursuing unilateral ASI supremacy through aggressive domestic innovation and Chinese compute containment, establishing a cooperative international consortium for shared development, or enforcing global technological suppression. Each approach carries profound operational trade-offs, ranging from triggering pre-emptive military strikes by Beijing over semiconductor access to crippling multilateral governance paralysis or squandering resources on uncertain capabilities. Because fully committing to any single option currently risks irrevocable failure, prudent hedging represents the optimal near-term strategy to preserve flexibility while safety benchmarks and technical parameters clarify.

Comment

Establishing multilateral oversight for superintelligence faces structural impediments far exceeding those that delayed the 1968 Nuclear Nonproliferation Treaty. Unlike nuclear material where centrifuges and reactors require physical footprint detection, dual-use compute clusters operate within decentralized digital architectures. Restricting technological diffusion to non-aligned states consequently demands unprecedented intrusive monitoring of commercial cloud providers and sovereign data centres. This requirement creates direct friction with European Allied regulatory standards on data sovereignty and industrial autonomy.

Strategic Question for Discussion

If monitoring dual-use compute clusters lacks the verifiable physical footprints inherent to the 1968 Nuclear Nonproliferation Treaty, which institutional mechanisms can prevent sovereign defectors from secretly training frontier models?

Share your assessment in the comments below.

To read in details click here