Overview
On November 1, 2023, representatives of 28 nations — including the United States, China, the European Union, and the United Kingdom — signed the Bletchley Declaration at the conclusion of the first AI Safety Summit, held at Bletchley Park in Buckinghamshire, England: the historic site where Alan Turing and colleagues broke Nazi Germany’s Enigma codes during World War II.
The Declaration was the first international governmental agreement specifically focused on the risks of frontier AI — the most advanced AI systems at the edge of current capabilities. Its significance lay not in its binding provisions (it had none) but in establishing that AI safety was a matter of shared international concern, and that even geopolitical rivals could find common ground on it.
What the Declaration Says
The Bletchley Declaration acknowledges that:
- Frontier AI models present “serious, even catastrophic” potential harms, including risks to critical infrastructure, biological and chemical weapons assistance, and loss of human control
- These risks are transnational — no single country can address them alone
- Transparency and accountability from AI developers are necessary conditions for safety
- International cooperation on AI safety research, standards, and information-sharing is essential
Notably, the Declaration avoided prescribing specific regulations, timelines, or enforcement mechanisms — a deliberate choice to achieve the unprecedented inclusion of China alongside Western democratic nations.
The Participants
The 28 signatories included:
- United States (represented by VP Kamala Harris, who arrived separately for the US-UK bilateral)
- China (represented by Vice Minister of Science and Technology Wu Zhaohui)
- European Union (represented by Commissioner Vera Jourová)
- UK, France, Germany, Japan, South Korea, India, Australia, Canada, Brazil, Nigeria, and others
China’s participation was diplomatically significant: it represented a rare area of collaboration between the US and China at a moment of acute geopolitical tension, and China’s subsequent contribution to the Seoul AI Summit (2024) suggested continued engagement.
The Frontier AI Framework
The Declaration introduced a shared vocabulary:
- Frontier AI: The most capable general-purpose AI models at the current technological edge
- Frontier risk: Risks arising specifically from the unprecedented capabilities of frontier models, distinct from narrower AI harms already covered by existing law
This distinction — between existing AI harms (bias, discrimination, privacy) and emergent frontier risks (autonomous deception, weapons uplift, loss of control) — shaped subsequent policy discussions globally.
Aftermath: Seoul and Beyond
The Bletchley Declaration established a rhythm of international AI safety summits:
- Seoul AI Summit (May 2024): 16 nations and leading AI companies signed the “Frontier AI Safety Commitments,” pledging safety testing and red-teaming before deploying frontier models
- Paris AI Action Summit (February 2025): Broader focus on AI governance, access, and economic impacts
- AI Safety Institutes Network: US, UK, EU, Japan, and others established coordinated AI safety evaluation bodies