AI Seoul Summit
Last reviewed
May 18, 2026
Sources
No citations yet
Review status
Needs citations
Revision
v1 · 3,515 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
May 18, 2026
Sources
No citations yet
Review status
Needs citations
Revision
v1 · 3,515 words
Add missing citations, update stale details, or suggest a clearer explanation.
The AI Seoul Summit was an international meeting on the safety, innovation and inclusivity of advanced artificial intelligence held on 21–22 May 2024 in Seoul, South Korea. Co-hosted by the Republic of Korea and the United Kingdom, it was the second instalment of the ai safety summit series inaugurated at Bletchley Park six months earlier, and the first to be held outside the United Kingdom.[^1][^2] The summit brought together heads of government, ministers, executives of leading frontier model developers, international organisations and civil society representatives across a hybrid virtual and in-person programme.[^1][^3]
This article concerns the summit event — the meeting itself, its participants, sessions and proceedings. The principal political document produced by leaders on Day 1, the Seoul Declaration for Safe, Innovative and Inclusive AI, is treated separately at seoul declaration, which sets out the declaration's text, signatories and substantive commitments.[^4] Other outputs covered in this article include the Seoul Statement of Intent toward International Cooperation on AI Safety Science, the Seoul Ministerial Statement, and the Frontier AI Safety Commitments signed by sixteen leading AI companies.[^2][^5][^6]
Although smaller in scale and media impact than its predecessor at Bletchley Park, the Seoul gathering is generally regarded by observers as a pivotal interim step that translated the high-level safety principles of the bletchley declaration into operational commitments — particularly the launch of a network of AI Safety Institutes and the first set of voluntary safety frameworks committed to by frontier developers.[^7][^8] The summit was succeeded by the paris ai action summit in February 2025 and the India AI Impact Summit in February 2026.[^9][^10]
| Item | Detail |
|---|---|
| Name | AI Seoul Summit (also "AI Seoul Summit 2024") |
| Dates | 21–22 May 2024 |
| Location | Seoul, Republic of Korea (Day 1 virtual; Day 2 in person) |
| Co-hosts | Republic of Korea and United Kingdom |
| Co-chairs (Leaders' Session) | President Yoon Suk Yeol (ROK); Prime Minister Rishi Sunak (UK) |
| Co-chairs (Ministerial) | Minister Lee Jong-Ho (ROK MSIT); Secretary of State Michelle Donelan (UK DSIT) |
| Predecessor | Bletchley Park, United Kingdom, 1–2 November 2023 |
| Successor | AI Action Summit, Paris, 10–11 February 2025 |
| Principal documents | Seoul Declaration; Seoul Statement of Intent; Seoul Ministerial Statement; Frontier AI Safety Commitments |
| Government signatories (Seoul Declaration) | 10 states + EU |
| Government signatories (Ministerial Statement) | 27 states + EU |
| Company signatories (Frontier AI Safety Commitments) | 16 frontier AI developers |
The Seoul gathering was the second event in a global series of high-level meetings on advanced AI risk that began with the ai safety summit hosted by the United Kingdom at Bletchley Park on 1–2 November 2023. The Bletchley meeting produced the bletchley declaration, in which 28 states and the European Union recognised that frontier AI models posed potentially catastrophic risks and committed to international cooperation on identification, evaluation and mitigation of those risks.[^11]
At the close of the Bletchley meeting, participants agreed in principle that follow-on summits would be held approximately every six months, with the Republic of Korea volunteering to host the second instalment and France the third.[^7][^11] The cadence — short interval, sustained political momentum — was intended to translate the principles laid out at Bletchley into more concrete deliverables: enhanced national capacity for model evaluation, voluntary commitments by developers, and an international scientific consensus on the state of AI risk.
Between Bletchley and Seoul, several developments shaped the agenda. The United States issued executive order 14110 on safe, secure and trustworthy AI in October 2023, and the US AI Safety Institute and UK AI Safety Institute were established and signed a bilateral cooperation memorandum on testing in April 2024.[^7][^12] The G7 Hiroshima AI Process produced its International Code of Conduct for organisations developing advanced AI systems, the OECD updated its AI Principles, and the EU AI Act moved toward final adoption.[^4][^13] These tracks competed for primacy in international AI governance, and one objective of the Seoul co-hosts was to position the AI Safety Summit series as a connective tissue between them rather than a rival forum.[^7][^4]
Within Korea, the summit was framed as an opportunity to demonstrate the country's status as a major AI producer (home to Samsung Electronics, Naver and other large developers) and to launch a domestic AI safety institute analogous to those in the UK and US.[^14][^7]
The interval between Bletchley and Seoul also saw rapid progress in frontier model capability. OpenAI's GPT-4 Turbo and Anthropic's Claude 3 family were released, and Google announced its Gemini 1.5 series. The first wave of pre-deployment safety evaluations conducted by the UK AI Safety Institute on these systems — and the institute's establishment of a research office in San Francisco — provided practical evidence both that government-led evaluation was technically feasible and that voluntary access depended on continuing developer goodwill, sharpening the case for the institutionalised cooperation framework that Seoul produced.[^7][^15]
The Seoul Summit adopted a hybrid format described by organisers and commentators as a "mini summit": a one-day virtual Leaders' Session followed by a one-day in-person Ministerial Session, rather than the two-day fully in-person meeting that had characterised Bletchley.[^15][^7] The structure was intended to maintain political momentum at a relatively short interval while reserving a larger physical convening for the subsequent Paris summit.
On the British side, the event was led by the Department for Science, Innovation and Technology (DSIT) and the UK AI Safety Institute; on the Korean side, by the Ministry of Science and ICT (MSIT) and other Korean government bodies.[^2][^14] Civil society side events were organised in Seoul through the "AI Global Forum" and a parallel "AI Fringe" series of meetings, though several civil society groups subsequently criticised what they characterised as limited substantive engagement of non-governmental voices in the official programme.[^16][^17]
The choice of a hybrid format had practical and political motivations. Practically, it allowed a much wider Day 2 invitation list without requiring all heads of government to travel to Seoul on short notice; politically, it underscored the message that the summit series would maintain a rapid cadence and a small core of leader-level decisions while expanding ministerial-level engagement.[^15][^7] The interval of approximately six months between Bletchley and Seoul became the de facto template — although later summits subsequently extended to an annual cadence under the broadened "AI Action" and "AI Impact" framings.[^9][^10]
The virtual Leaders' Session was attended by representatives of ten states plus the European Union: Australia, Canada, France, Germany, Italy, Japan, the Republic of Korea, the Republic of Singapore, the United Kingdom and the United States, together with the EU.[^4][^18] These were the parties that adopted the Seoul Declaration for Safe, Innovative and Inclusive AI and the Seoul Statement of Intent toward International Cooperation on AI Safety Science.[^4][^5]
The session was co-chaired by President Yoon Suk Yeol of the Republic of Korea and Prime Minister Rishi Sunak of the United Kingdom.[^15][^18] A selected group of frontier AI company executives joined parts of the leaders' programme to report on how they had been fulfilling the safety commitments made at Bletchley and to announce the Frontier AI Safety Commitments.[^15][^6]
The in-person Ministerial Session expanded the participant set considerably. It was co-chaired by UK Secretary of State for Science, Innovation and Technology Michelle Donelan and Korean Minister of Science and ICT Lee Jong-Ho.[^15] In addition to the eleven Leaders' Session parties, the Ministerial drew representatives from Chile, India, Indonesia, Israel, Kenya, Mexico, the Netherlands, New Zealand, Nigeria, the Philippines, Rwanda, Saudi Arabia, Spain, Switzerland, Türkiye, Ukraine and the United Arab Emirates, producing a Seoul Ministerial Statement signed by 27 states plus the EU.[^2][^19] China participated in summit discussions but did not sign the Ministerial Statement.[^13][^19]
International organisations represented included the United Nations, the Organisation for Economic Co-operation and Development (OECD), and the European Union.[^18][^2]
Industry representation was led by the chief executives or senior technical leaders of the principal frontier AI developers. Reported attendees included Elon Musk of Tesla and xAI, Lee Jae-yong of Samsung Electronics, and senior representatives of OpenAI, Google DeepMind, Microsoft, Meta and Naver, among others.[^18] Civil society organisations and academic researchers contributed primarily through side events and an open consultation process, although their presence at the formal sessions was limited compared with Bletchley.[^16][^17]
The virtual Leaders' Session on 21 May 2024 was organised around three thematic pillars established jointly by the co-hosts: safety, innovation, and inclusivity. The addition of innovation and inclusivity alongside safety was a deliberate broadening relative to Bletchley, reflecting the Korean co-host's emphasis on AI as a tool of economic development and the UK's interest in linking safety to continued frontier research.[^4][^7]
The session produced three principal documents on Day 1:
In parallel, the UK government and the Alan Turing Institute announced £8.5 million in research funding for "systemic AI safety" research, and the shortlist for the UK's £1 million Manchester Prize for AI for public benefit was published.[^3][^15]
The Leaders' Session was preceded on 17 May 2024 by the publication of the interim International Scientific Report on the Safety of Advanced AI, chaired by Yoshua Bengio and prepared by an international expert panel under the Bletchley follow-up process. The interim report — later renamed the International AI Safety Report in its final 2025 version — provided a shared scientific baseline for the political discussions that took place at Seoul, summarising current and anticipated frontier-AI capabilities, plausible risk pathways, and the state of evaluation and mitigation techniques.[^1][^19]
The in-person Ministerial Session on 22 May 2024 was held in Seoul and chaired by Ministers Donelan and Lee. Its principal output was the Seoul Ministerial Statement, agreed by 27 states plus the EU, which set out commitments to develop proposals for assessing risks of advanced AI — including shared thresholds for "severe risks" such as misuse for biological, chemical, radiological or nuclear weapons development, and risks of AI safety failure including loss of human oversight — for consideration at the next summit.[^19][^2]
The Ministerial Session also addressed AI's environmental impact (including a discussion of low-power AI chips), workforce effects, equitable benefit distribution, and the coordination of national AI governance regimes. South Korea's plan to establish its own AI Safety Institute was formally announced during the session.[^14][^2]
The Seoul Declaration restated and extended the principles of the bletchley declaration. It articulated a shared commitment to "safe, innovative and inclusive" AI, recognised interoperability between national and international AI governance frameworks as essential, and explicitly linked the Summit series with the G7 Hiroshima AI Process Code of Conduct, the OECD AI Principles and the UN's AI work. Full text and analysis appear at seoul declaration.[^4]
The Seoul Statement of Intent toward International Cooperation on AI Safety Science was an annex to the Declaration signed by the same eleven parties.[^5] It committed signatories to coordinate on the science of frontier AI safety through their AI Safety Institutes or "similar institutions". Specific mechanisms identified included:
Operationally, the Statement of Intent provided the foundation for the International Network of AI Safety Institutes, formally launched in San Francisco in November 2024 and meeting subsequently at Paris in February 2025.[^20] Members included the UK AI Safety Institute, US AI Safety Institute (renamed the Center for AI Standards and Innovation in 2025), the new Korean AI Safety Institute, and the corresponding bodies of Australia, Canada, France, Germany, Italy, Japan, Singapore and the EU AI Office.[^20][^5]
A central deliverable of Day 1 was the Frontier AI Safety Commitments, in which sixteen leading developers of frontier AI systems committed to publish, before the next summit, organisational frameworks specifying how they would identify, assess and mitigate severe risks from their frontier models.[^6] The initial signatories were:
Four additional firms — Magic, MiniMax, 01.AI and NVIDIA — joined the commitments in the period leading up to and following the Paris summit.[^6]
The commitments are organised around three "outcomes":
In an unprecedented step, signatories committed that, where mitigations cannot keep risks below their defined thresholds, they would "not develop or deploy [the] model or system at all".[^6][^21] The frameworks were intended to be published by the time of the next summit, and many — including Anthropic's Responsible Scaling Policy, Google DeepMind's Frontier Safety Framework, and OpenAI's Preparedness Framework — were updated or released to satisfy the Seoul commitment.[^21]
The Ministerial Statement signed on Day 2 was a broader document agreed by 27 states plus the EU. It committed signatories to work together on identifying severe AI risks (including biosecurity, cybersecurity and loss of human oversight) and on developing proposals for risk thresholds for consideration at the Paris summit. It also addressed workforce, sustainability and equitable benefit-sharing themes and welcomed the publication on 17 May 2024 of the interim International AI Safety Report chaired by Yoshua Bengio.[^2][^19]
Alongside the principal documents, several states used the summit to make national commitments. The Republic of Korea announced the establishment of its AI Safety Institute (later launched in November 2024) and a domestic AI Framework Act. The UK announced £8.5 million in systemic AI safety research funding and the opening of a UK AI Safety Institute office in San Francisco.[^15][^14] Canada announced new AI investment programmes including funding for its AI Safety Institute.[^22]
Initial reception of the AI Seoul Summit was mixed. Industry analysts and policymakers broadly welcomed the Frontier AI Safety Commitments and the formal launch of the AI Safety Institutes' coordination as concrete, operationalisable outcomes that built directly on Bletchley.[^8][^7] The Center for Strategic and International Studies (CSIS) characterised the institutes network as "the moment that the idea reached significant international scale", arguing that the summit's substantive achievements were obscured by lower media visibility relative to Bletchley.[^7]
Critics, however, expressed several concerns:
The Centre for Emerging Technology and Security at the Alan Turing Institute, in a "stocktake" published after the summit, concluded that Seoul had successfully translated Bletchley's principles into concrete deliverables but that the next summit would need to demonstrate measurable risk-reduction outcomes rather than further framework-building.[^24]
The Seoul Summit handed the AI Safety Summit baton to France, which had volunteered at Bletchley to host the third instalment. Two further summits in the series have since taken place:
The Frontier AI Safety Commitments framework launched at Seoul has continued to expand at subsequent summits, with additional companies acceding and the original signatories publishing successive iterations of their safety frameworks. As of 2026, the Commitments are widely regarded as the institutional ancestor of company-level safety-framework regimes such as the Responsible Scaling Policy, Frontier Safety Framework and Preparedness Framework.[^21][^6]