The future according to AI

A new state declares or tests a nuclear weapon and the non-proliferation bargain visibly fails

The taboo on new entrants breaks. A state either withdraws from the treaty under its withdrawal clause, conducts a detectable test, or officially declares…

Claude · 2032–2042 · grey-swan

Prior state

The last bilateral strategic arms limitation between the largest arsenals lapsed in 2026 with no successor. Several states possess the technical capacity to build a weapon within a short period and have refrained on the basis of alliance guarantees: in one case a clear majority of the public has favoured an indigenous capability for years. Another state's programme has been damaged by military action but not eliminated, and its inspection arrangements are contested. The review cycle of the non-proliferation treaty has produced no consensus outcome for over a decade.

Material change

The taboo on new entrants breaks. A state either withdraws from the treaty under its withdrawal clause, conducts a detectable test, or officially declares a capability. The treaty's claim to near-universality ends, and the central question shifts from prevention to whether a regional cascade follows.

Why now

The restraining variable is the credibility of extended deterrence, and the decade contains multiple leadership transitions in guarantor states, at least one treaty review cycle, and the semiconductor and alliance shifts described above, each of which affects that credibility. Latent states can move from decision to capability in a period much shorter than ten years, so the constraint is political will and its supply is what changes.

Mechanism and resistance

Alliance reassurance, sanctions, safeguards and domestic opposition have all worked for five decades and may work again. The decision is enormously costly economically and diplomatically, and the states most capable of taking it are also the most exposed to the sanctions that would follow. Testing is detectable, which is why a declared-but-untested capability is a plausible intermediate form. Suppliers of enrichment and reprocessing technology retain real leverage.

Consequences

The most important consequence is sequencing: a single entrant creates immediate pressure on its neighbours, and the states in question are concentrated in two regions where several have latent capacity. Sanctions architecture strains because the entrant may be an ally of the states expected to enforce them. Missile-defence and counterforce spending rises across both regions. The verification and safeguards system, which was built for a universal regime, must be rebuilt for a partial one, and the institutional work of doing so occupies the rest of the decade and beyond.

End state

Either a newly declared nuclear state and a non-proliferation regime that survives formally while having lost its central claim, or a regime that is visibly hollow and held together only by alliance guarantees whose credibility is openly debated.

Observable test

A detected test recorded by the international monitoring system, a formal withdrawal notification under the treaty's withdrawal article, or an official state declaration of possession; changes in safeguards status reported by the international agency.

Disconfirming sign

The review cycle passes with no withdrawal notification, and latent states publicly reaffirm reliance on alliance-based deterrence.

Themes

Security & cyber, Geopolitics, Law & institutions