Britain's top diplomat just put AI in the same threat category as nuclear weapons — and she's betting the farm on getting Trump and Xi to shake hands over it.
The Summary
- UK Foreign Secretary Yvette Cooper called AI a "Hiroshima"-style existential risk requiring global coordination between the US and China within two years
- Cooper frames AI alongside climate crisis, irregular migration, and foreign interference as interlinked threats that demand multilateral action
- The UK is positioning itself as the broker for AI governance talks between rival superpowers, despite having limited enforcement power
The Signal
Cooper's framing matters more than her proposed solution. She's packaging AI with climate and migration, three problems that share one trait: they don't respect borders. That bundling tells you where UK foreign policy is headed. Expect AI governance to get tangled up with carbon credits and refugee quotas at every G7 meeting for the foreseeable future.
The "Hiroshima" comparison is deliberate provocation. Nuclear weapons got the Non-Proliferation Treaty because five countries held all the cards and had recent memory of what happens when you don't talk. AI is different. The code is already out there. Thousands of labs across dozens of countries can train frontier models. Cooper's call for US-China cooperation assumes both sides want the same thing enough to compromise. That's not obvious.
"The issue will dominate foreign policy over the next two years."
What Cooper isn't saying: the UK has almost no leverage here. Britain isn't building the models that matter. DeepMind is owned by Google. Anthropic is American. The Chinese labs don't take calls from Westminster. The UK's play is to host summits and draft frameworks, hoping moral authority translates to influence. It worked for nuclear arms control when Britain still had bombs. For AI, they're bringing a white paper to a compute war.
Key gaps in Cooper's position:
- No proposed mechanism for verification or enforcement
- No acknowledgment that AI development happens in private labs, not government facilities
- No plan for what happens when one country defects from voluntary agreements
The two-year timeline is telling. That's one election cycle in the US, maybe two government reshuffles in the UK. Cooper is betting that the current window of public concern about AI safety stays open long enough to lock in treaties. History suggests otherwise. The urgency fades. The lobbyists arrive. The frameworks get watered down until they regulate nothing that matters.
The Implication
Watch for the UK to push hard for AI governance language in every multilateral agreement through 2027. Climate summits, trade pacts, security forums — they'll all get AI clauses. Most will be unenforceable theater. The real action stays where it's always been: inside the companies training the models and the governments funding them. If you're building agents or infrastructure for Web4, don't assume global rules will slow you down. Assume they'll create compliance theater that big players can afford and small players can't.