UK Foreign Secretary Yvette Cooper argues in a Chatham House essay that Britain should lead a global AI safety accord, built before any 'AI equivalent of Hiroshima' forces the world's hand.
UK Foreign Secretary Yvette Cooper is telling world governments to agree on international AI safety rules now, before catastrophe forces their hand. In an essay set to publish Monday at Chatham House, the international affairs think tank, she argues there is no global equivalent of the post-1945 nuclear accords for AI, and that the world "cannot afford to wait for an AI equivalent of Hiroshima before we act" (Penarth Times).
Cooper is positioning the United Kingdom as the country best placed to lead the work, citing the 2023 AI Safety Summit at Bletchley Park, the WWII codebreaking site that then-prime minister Rishi Sunak turned into a two-day gathering of governments, frontier AI labs, and civil society. Her argument is structural: the world has a coordination gap with a clear historical precedent, and Britain wants to be the convener of the next round.
That framing lands the same week the UN Independent International Scientific Panel on AI released its preliminary report, which warned that the gap between rapidly improving AI capabilities and effective risk-management methods "may lead to catastrophic outcomes" (UN Independent International Scientific Panel on AI). UN Secretary-General António Guterres framed the assessment at the panel launch as "the science is here" (UN News). The panel is the first product of the body created at the Bletchley Park summit and ratified since.
Cooper's urgency also leans on the Five Eyes intelligence alliance, the UK, US, Canada, Australia, and New Zealand, whose warning that AI-powered cyber attacks could be just months away is cited in her essay. Coverage of the UN panel has echoed the gap framing: Business Standard reported the panel's "catastrophic risks" warning (Business Standard), and Allwork.Space summarized the same finding for a workplace-policy audience (Allwork.Space).
The Hiroshima analogy is Cooper's quoted hook, not the essay's thesis. Her core argument is that after the United States dropped atomic weapons on Hiroshima and Nagasaki in 1945, the world's major powers built arms-control, non-proliferation, and test-ban treaties over decades because they had seen what the technology could do. AI, she argues, has no equivalent of the Baruch Plan, the Nuclear Non-Proliferation Treaty, or the IAEA. There is no standing international body with inspection authority over frontier AI labs, no baseline rule on what counts as a prohibited capability, and no common red lines that all major compute-owning jurisdictions accept.
Whether Britain can credibly lead that work is the open question. Cooper's domestic baseline is the March 2023 UK White Paper "A pro-innovation approach to AI regulation," which deliberately avoided hard rules in favour of letting existing regulators adapt their sectors (UK Government). The essay does not disown that document; it argues the safety conversation has moved past it. Cooper is pitching a UK role in international coordination while her own government's domestic posture remains light-touch, a tension the available excerpts of her essay do not directly resolve.
The Chatham House publication on 2026-07-06 is the next test. Watch whether Cooper names a specific forum (a second AI safety summit, a G7 track, a UN process), commits UK funding, or anchors the Five Eyes cyber warning in a joint statement with a binding cyber-AI pledge.