one year on
28 Countries and EU Sign Bletchley Declaration on Frontier AI Risks
At the UK's AI Safety Summit, 28 countries and the European Union agree to cooperate on identifying and mitigating dangers from the most advanced AI models.
The UK’s AI Safety Summit, held at Bletchley Park—the historic site of World War II codebreaking—brings together representatives from 28 countries and the European Union to sign the Bletchley Declaration on the risks posed by frontier AI.
The declaration, dated November 1, 2023, acknowledges that AI systems ‘already deployed across many domains of daily life’ and that ‘particular safety risks arise at the frontier’ of AI, including highly capable general-purpose models. The signatories—which include both the United States and China—agree to intensify cooperation to identify, understand, and act against ‘serious, even catastrophic’ harms from misuse or loss of control. The declaration notes that ‘substantial risks may arise from potential intentional misuse or unintended issues of control relating to alignment with human intent’ and that these risks are ‘inherently international in nature, and so are best addressed through international cooperation.’ The signatories committed to meet again in 2024.
The record
One year later — open only if you can handle spoilers
The declaration’s broad participation was widely seen as diplomatic success, but critics noted it lacked enforcement mechanisms. A year later, several signatories had introduced divergent domestic regulations, while frontier AI capabilities continued to advance without binding international rules.