The Bletchley Declaration and What Came After

---

The Setting: Bletchley Park

The choice of Bletchley Park as the venue for the AI Safety Summit was deliberate and symbolic — connecting the new technology of AI to the long history of computing. The summit was hosted by UK Prime Minister Rishi Sunak, who had made AI safety a priority of his government. It was held on November 1 and 2, 2023, and brought together representatives from 28 countries, the European Union, leaders of major AI companies, and civil society organisations.

The most notable attendee, from a geopolitical perspective, was Wu Zhaohui, China’s Vice Minister of Science and Technology. China’s participation was significant because the United States and China are the two major AI powers. The decision to invite China was controversial in some quarters, but Sunak defended it, arguing that any serious strategy for AI safety had to engage all of the world’s leading AI powers. The summit also featured Sam Altman (OpenAI — who had testified before the Senate six months earlier), Demis Hassabis (Google DeepMind), and Elon Musk (xAI), with whom Sunak held a public conversation on the second day.

---

The Bletchley Declaration

The Bletchley Declaration was signed on November 1, 2023. It is a short document — about two pages — and it is not a binding treaty. It is a statement of concern and a commitment to future cooperation.

The declaration focuses on “frontier AI” — the most advanced AI models, at the frontier of capability — which it says could cause “serious, even catastrophic, harm.” The key commitments are three: (1) the signatories agree to support an internationally inclusive network of scientific research on frontier AI safety; (2) they agree to work towards identifying AI safety risks of shared concern and building a shared scientific understanding; (3) they agree to meet again. The declaration does not establish binding rules, enforcement mechanisms, or specific regulatory approaches.

---

The UK AI Safety Institute

On October 26, 2023, Sunak announced the creation of the UK AI Safety Institute (AISI) — the world’s first state-backed AI safety institute — which was formally launched at the summit on November 1–2, 2023. The institute’s mission was to examine, evaluate, and test new types of frontier AI models. On February 14, 2025, the UK AISI was renamed the AI Security Institute, with a shift in emphasis from “safety” (risks to the public) to “security” (use by hostile states, cybersecurity, protection of critical infrastructure).

---

The US AI Safety Institute

On October 30, 2023 — two days before the Bletchley summit — President Biden signed Executive Order 14110, “Safe, Secure, and Trustworthy Development and Use of AI,” which directed NIST (the National Institute of Standards and Technology) to establish AI safety work. The US AI Safety Institute (USAISI) was founded within NIST in November 2023. In February 2024, Elizabeth Kelly was named its inaugural director, and NIST also established the AI Safety Institute Consortium (AISIC) with over 200 member organisations (NIST cited 280+ at launch). Japan launched its own AI Safety Institute (J-AISI) in February 2024 as well.

---

The Seoul Summit: May 2024

The second major summit was held on May 21–22, 2024, in Seoul, co-hosted by the UK and South Korea (virtual leaders’ session May 21; in-person ministerial May 22). The Seoul Summit produced the Frontier AI Safety Commitments — signed by 16 AI companies (Amazon, Anthropic, Cohere, Google/DeepMind, G42, IBM, Inflection AI, Meta, Microsoft, Mistral AI, Naver, OpenAI, Samsung, and others). The companies agreed to publish safety frameworks, define “red lines” beyond which AI development would not proceed without safety assurances, and allow external testing of their models.

---

The Paris AI Action Summit: February 2025

The third major summit was held on February 10–11, 2025, in Paris, co-chaired by France and India. It was billed as an “AI Action Summit” — a shift in emphasis from the “safety” focus of Bletchley and Seoul to a broader focus on “action,” including economic opportunity, public-interest AI, and innovation.

Emmanuel Macron announced €109 billion (~$113 billion) in private-sector AI investments in France — a major commitment designed to position France and Europe as significant players in the AI industry. The summit also launched Current AI, a public-interest AI foundation with an initial endowment of about $400 million (backed by nine governments), designed to support AI systems serving the public interest. The summit’s final declaration — “Statement on Inclusive and Sustainable AI for People and the Planet” — was signed by about 60 nations and organisations, including China, India, France, Japan, Australia, and Canada.

But the most notable thing about the Paris Summit was who did not sign: the United States and the United Kingdom both declined.

---

JD Vance and the American Pivot

The US decision not to sign the Paris declaration was the result of a significant shift in American AI policy under the new Trump administration. On January 20, 2025, the Trump administration rescinded Biden’s Executive Order 14110 (via EO 14148). On January 23, 2025, Trump signed EO 14179, “Removing Barriers to American Leadership in AI.”

JD Vance articulated the new approach at the Paris Summit on February 11, 2025 — his first foreign trip as Vice President. He said: “We believe that excessive regulation of the AI sector could kill a transformative sector just as it’s taking off, and we’ll make every effort to encourage pro-growth AI policies.” He also said: “I’m not here this morning to talk about AI safety, which has been a focus of a number of prior AI summits. I’m here to talk about AI opportunity.”

This was a significant shift: from “AI safety” as the primary concern to “AI opportunity” as the primary concern, with “excessive regulation” framed as the bigger threat. The UK also declined to sign the Paris declaration, though for somewhat different reasons (the UK government said it did not go far enough on governance and national security).

Later in 2025, the Trump administration rebranded USAISI as the Center for AI Standards and Innovation (CAISI) — a shift from “safety” to “standards and innovation.” The administration’s Department of Government Efficiency (DOGE) also targeted NIST for significant cuts, with as many as 500 NIST employees reportedly facing dismissal, affecting the AI safety institute’s capacity.

---

What the Bletchley Process Achieved

Despite the setbacks, the Bletchley process achieved several things. The achievements were primarily institutional and procedural, building the architecture for international cooperation on AI even where the substantive rules were missing. The most important gains were the following.

1. It established AI safety as a subject of international diplomatic attention — a topic that heads of state discussed at summits.
2. It created a network of national AI safety institutes (UK, US, Japan, and others) that could share research and coordinate on safety standards.
3. It produced voluntary commitments from the major AI companies (the Frontier AI Safety Commitments at Seoul).
4. It established a pattern of regular summits (Bletchley → Seoul → Paris → India 2026).

What the Bletchley Process Did Not Achieve

The achievements, however, were more procedural than substantive, and several fundamental problems remained unresolved. The Bletchley process did not produce binding rules, did not resolve the central tensions in the field, and did not include much of the world. The most important gaps were as follows.

1. It did not produce binding rules. There is still no binding international agreement on AI safety.
2. It did not resolve the fundamental tension between safety and innovation. The Bletchley/Seoul summits emphasised safety; the Paris summit and the Trump administration emphasised innovation.
3. It did not address the geopolitical competition over AI (US vs. China).
4. It did not meaningfully include the Global South. The India AI Impact Summit, scheduled for February 2026 in New Delhi, will be the first global AI summit in the Global South.
5. It did not resolve the question of what “AI safety” actually means — the UK’s shift from “safety” to “security” and the US shift from “safety” to “standards and innovation” both reflect a narrowing or redefinition of the concept.

---

Where Things Stand

The international AI governance process that began at Bletchley Park in November 2023 is still ongoing, but it has entered a more contested phase. The early consensus — that AI safety required urgent international cooperation — has fractured. The US, under the Trump administration, has signalled that it sees excessive regulation as a bigger threat than AI. The UK has narrowed its focus from safety to security. The EU has pressed ahead with its own AI Act.

The next major test will be the India AI Impact Summit in February 2026 in New Delhi — the first global AI summit in the Global South, and a test of whether the international process can survive the shift in US priorities.

The Bletchley Declaration was a beginning, not an end. It established AI safety as a subject of international attention, and it created the institutions and the patterns of cooperation that will shape the international debate for years to come. But the hard work — the work of turning statements of concern into binding rules, of balancing safety with innovation, of engaging the Global South, of cooperating despite geopolitical competition — is still ahead. The Bletchley process was the first chapter of the international governance of AI. The next chapters are still being written.

---

---

Was the bletchley declaration inevitable — the product of forces too large to redirect — or was it a series of choices, each of which could have gone differently? The answer matters, because it determines whether the future is something that happens to us or something we make.