First global AI Safety Summit adopts the Bletchley Declaration

A first global meeting on frontier AI

The United Kingdom convened the first global AI Safety Summit at Bletchley Park on 1 and 2 November 2023^[1]. The meeting centred on what the declaration calls frontier AI: the most capable models of a given moment, whose behaviour is difficult to fully specify or test in advance. Its main output was a political declaration — a statement of shared understanding among governments rather than a set of binding technical rules.

What the Bletchley Declaration says

On 1 November 2023, twenty-eight countries, among them the United States, China and the European Union, agreed the Bletchley Declaration^[1]. The document established a shared understanding of the opportunities and risks posed by frontier AI^[1], setting out both the potential benefits of the technology and the dangers that governments judged to require attention.

On risk, the signatories agreed that substantial risks may arise from two broad channels: potential intentional misuse of frontier AI, and unintended issues of control^[1]. They named cybersecurity, biotechnology and disinformation as areas of particular concern^[1]. The declaration further stated that, stemming from the most significant capabilities of these models, there is potential for serious, even catastrophic, harm, whether deliberate or unintentional^[1][2].

The declaration also set a direction for how governments would try to understand those risks. Countries resolved to support an internationally inclusive network of scientific research on frontier AI safety^[2] — an effort framed around shared scientific work rather than unilateral national assessment alone.

Commitments beyond the text

The summit produced a schedule for continuing the conversation. The Republic of Korea agreed to co-host a mini virtual summit within six months, and France agreed to host the next in-person summit in a year^[1].

Institutions were announced alongside the declaration. The UK announced that it would establish the world's first AI Safety Institute^[1]. It also commissioned Yoshua Bengio to chair an international "State of the Science" report on the capabilities and risks of frontier AI^[3], intended to consolidate what is known about those capabilities and risks across the research community.

The declaration remained open to further signatories after the summit. New Zealand joined the commitment to the Bletchley Declaration on 23 October 2024^[2], close to a year after the meeting itself.

Why it matters

The summit's significance lies less in any single commitment than in the fact of agreement. Governments that compete on AI development — including the United States, China and the European Union — put their names to a common description of what could go wrong with frontier systems, and to a shared list of concerns: misuse, loss of control, cyber and biological risk, and disinformation^[1]. That gave later international discussion a common vocabulary.

The meeting also left behind working machinery rather than only a statement: a named national institute, an internationally chaired scientific report, a proposed research network, and a calendar of further summits in Korea and France^[1][3][2]. Read together, these are mechanisms for producing and sharing evidence about model capabilities and risks — the part of AI governance that has to be built before rules can be written.

The declaration is explicit about uncertainty as well as danger: it speaks of risks that may arise from misuse or from unintended issues of control, and of potential for serious harm from the most significant capabilities of these models^[1][2]. That framing — precaution without a specific ban, cooperation without a treaty — is what the first global AI safety summit actually agreed.