
World Artificial Intelligence Cooperation Organization
The World Artificial Intelligence Cooperation Organization (WAICO) is an independent intergovernmental organization with international legal personality…
Expert article
The World Artificial Intelligence Cooperation Organization (WAICO) is an independent intergovernmental organization with international legal personality, headquartered in Shanghai, for cooperation on artificial intelligence. Its agreement states four purposes: a human-centric approach promoting AI "for good and for all"; an open, fair and non-discriminatory environment for development; bridging an "AI divide" and widening access to AI technologies and services; and collective work to mitigate risks and to promote safe, secure, ethical and trustworthy AI. The functions, confined to civilian uses, include a platform for matching supply and demand among members, capacity-building, coordination of development strategies, governance rules and technical standards, scientific cooperation and an open-source ecosystem, support for governance processes inside the United Nations framework, and contact with other international organizations. The organization is to follow the purposes and principles of the Charter of the United Nations and to work through a council of all member states and a secretariat headed by a secretary-general.
On 16 July 2026, the day before the 2026 World Artificial Intelligence Conference opened in Shanghai, officials from 29 countries signed the agreement. Secretary-General António Guterres attended, as did deputy prime ministers of Kazakhstan, Laos and Pakistan, and Wang Yi, foreign minister of China, signed for the Chinese government. Al Jazeera named Indonesia, Brazil, Malaysia, South Africa, Senegal, Russia and Pakistan among the founding group. The Straits Times reported that the ASEAN members Cambodia, Indonesia, Laos, Malaysia and Myanmar were among the 29. The Diplomat described the signatories as countries from Asia, Africa and Latin America and wrote that no Western or strongly Western-aligned country was among the founders. Under the text, the agreement was open for signature from 1 May to 31 July 2026, founding members are the states that sign in that window, membership is open to all states, and the agreement enters into force when a third instrument of ratification, acceptance or approval is deposited with the Chinese government, which is the depositary.
Li Qiang, the Chinese premier, proposed a global AI cooperation organization on 26 July 2025 at the opening of that year's World AI Conference in Shanghai, and said AI should become an international public good and remain under human control. He said China was willing to share development experience and open-source technology, especially with countries in the Global South, and an action plan for global AI governance was issued at the same meeting. In a May 2026 paper for the Carnegie Endowment for International Peace, Arindrajit Basu described that proposal as short on administrative detail but as evidence that Beijing meant to shape AI governance institutions, complement its work at the United Nations, and expand a network of AI cooperation centres. On 1 November 2025, at an Asia-Pacific Economic Cooperation leaders' meeting, Xi Jinping said China had proposed the organization so that it would supply "public goods on AI" through cooperation on development strategies, governance rules and technological standards, and that China would work with APEC members on AI literacy and the regional digital divide. In December 2025, Nature reported that China, having placed AI at the centre of its economic strategy, was pressing for an international system to govern the technology while binding global rules were still absent.
The day after the signing, Xi told the Shanghai conference that AI development "should not be a solo performance by a single country, but a symphony of international cooperation," and that states should oppose "overstretching the national security concept" in the field. He also called for a people-centred approach in which AI stayed under human control through regulation, technological monitoring, early warning and emergency response. The Diplomat reported that he announced plans for 5,000 training places for developing countries over five years and for cooperation centres with groups including ASEAN, the African Union and BRICS, and wrote that it remained to be seen whether the signing would be followed by working groups, shared projects and funding. In September 2026, Josephine Teo, Singapore's minister for digital development and information, told Parliament that Singapore had been invited to join and was still considering the invitation, describing the body as a counterpoint to Pax Silica, a United States-led initiative on AI supply chains that Singapore had joined in December 2025. Al Jazeera reported that analysts expected Beijing to use the organization to influence how AI policy is framed at the United Nations, and the BBC later described the July founding as part of China's effort to shape global AI governance in its rivalry with the United States.
Establishment
Proposals in 2025
China proposed a standing intergovernmental body on artificial intelligence at the 2025 World Artificial Intelligence Conference in Shanghai. On 26 July, Premier Li Qiang said Beijing wanted an organization to coordinate rules and share Chinese advances, warning that AI risked becoming an "exclusive game" of a few countries and companies. He described global governance as fragmented and called for a consensus framework, offering to share experience and products especially with the Global South. The remarks followed an export-oriented AI blueprint from the administration of U.S. President Donald Trump. Vice Foreign Minister Ma Zhaoxu told officials from more than 30 countries, including Russia, South Africa, Qatar, South Korea and Germany, that China wanted practical cooperation and was considering Shanghai as the seat. The foreign ministry also issued an action plan that invited governments, organizations, companies and researchers to cooperate, including through a cross-border open-source community.
Xi Jinping restated the proposal on 1 November 2025 at an Asia-Pacific Economic Cooperation leaders' meeting in Gyeongju, South Korea. Reuters described the remarks as his first public comments on the initiative. In comments carried by Xinhua, he said a World Artificial Intelligence Cooperation Organization could set governance rules and make AI a "public good for the international community." Officials again indicated Shanghai as a possible headquarters. The same account said the United States had rejected regulating AI in international bodies. Trump did not attend the leaders' session, and APEC members adopted a joint declaration and separate pacts on AI.
Signing ceremony
On 16 July 2026, the eve of the next conference, representatives of 29 countries signed the Agreement on the Establishment of the World Artificial Intelligence Cooperation Organization in Shanghai. Foreign Minister Wang Yi signed for China. The foreign ministry named Kazakhstan's deputy prime minister Zhaslan Madiyev, Laos's deputy prime minister Thongsavanh Phomvihane, Pakistan's deputy prime minister Mohammad Ishaq Dar, Russian official Maxim Oreshkin and Indonesia's coordinating minister Airlangga Hartarto among the signatories. United Nations Secretary-General António Guterres attended but was not reported to have signed. Al Jazeera identified Indonesia, Brazil, Malaysia, South Africa, Senegal, Russia and Pakistan among the 29, and a Maseru law firm reported that foreign minister Limpho Tau signed for Lesotho. The governments named in that coverage lay mainly outside the largest Western economies.
On 17 July, Xi opened the 2026 World Artificial Intelligence Conference and called the organization a "milestone in the history of world AI development." He said development "should not be a solo performance by a single country, but a symphony of international cooperation," urged states to oppose "overstretching the national security concept," and argued that systems should remain under human control. Al Jazeera read the warning against domination by one country as directed at the United States. China's government news service reported pledges of 5,000 training places for developing countries over five years and of application centres with ASEAN, the League of Arab States, the African Union, the Community of Latin American and Caribbean States, the Shanghai Cooperation Organisation and BRICS. Guterres, Kazakh president Kassym-Jomart Tokayev and Thai prime minister Anutin Charnvirakul were reported at the opening; a group photograph also showed Cambodian prime minister Hun Manet.
The text creates an independent intergovernmental organization with international legal personality, a Shanghai headquarters and a civilian-only mandate, and it requires the body to follow the purposes and principles of the U.N. Charter. States that sign during the open period are founding members; other states join only by application and council-approved accession.
Signature window and later accessions
Article 14 kept the agreement open for signature from 1 May until 31 July 2026, subject to ratification, acceptance or approval, and left later accession available only after council approval. Eight further governments signed before the close. On 30 July, Dominica's ambassador Martin Charles signed. On 31 July, Assistant Foreign Minister Liu Bin witnessed signatures for Brunei, Togo, Iran, Sudan and Vietnam—by minister Mohammed Riza, minister Cina Lawson, ambassador Abdolreza Rahmani Fazli, ambassador Omer Eisa Ahmed and chargé d'affaires Hoang Van Tuan—while Georgia's ambassador Paata Kalandadze and Tanzania's ambassador Suleiman Haji Suleiman also signed. With the 29 of 16 July, the reported total was 37. Those signatures still fall inside the founding-member window rather than the accession procedure.
Article 15 brings the agreement into force only when the third instrument of ratification, acceptance or approval is deposited with China, the depositary, and for each later state on the date of its own deposit. A Lesotho firm wrote on 27 July 2026 that signature was not yet a change in domestic law. On 24 August, Konstantinos D. Magliveras noted that three ratifications could let the body start with very few members, and he treated further ratifications as still unscheduled. That analysis, the latest dated commentary among the sources opened here, did not report a depositary notice that a third instrument had been deposited or that the agreement had entered into force.
Membership
Founding signatories
The agreement opens membership to all states. Those that sign while it is open for signature, from 1 May through 31 July 2026, are founding members; any other state must apply through the secretariat and may accede only with Council approval after that window. Signature still requires ratification, acceptance or approval. The agreement enters into force when the third such instrument is deposited with the government of China, the depositary, and for each later state on the date of its own deposit. An August 2026 commentary treated those signatures as still short of ratification, while noting that three deposits would be enough for entry into force.
On 16 July 2026 in Shanghai, representatives of 29 countries signed. The Chinese foreign ministry named, among the representatives, Kazakhstan’s Deputy Prime Minister and artificial-intelligence minister Zhaslan Madiyev, Lao Deputy Prime Minister Thongsavanh Phomvihane, Pakistan’s Deputy Prime Minister and Foreign Minister Mohammad Ishaq Dar, Russian presidential official Maxim Oreshkin, and Indonesia’s Coordinating Minister for Economic Affairs Airlangga Hartarto. United Nations Secretary-General António Guterres attended; the United Nations did not sign.
An August annex from the EU-funded SESEC project and an enumerated list published via Forbes África Lusófona name the same 29 states: Algeria, Belarus, Brazil, Cambodia, Cameroon, China, Congo, Cuba, Ethiopia, Indonesia, Kazakhstan, Kenya, Kyrgyzstan, Laos, Lesotho, Malaysia, Mozambique, Myanmar, Nicaragua, Oman, Pakistan, Russia, Senegal, Serbia, South Africa, Tajikistan, Uzbekistan, Venezuela and Zambia. Separate reporting identifies that “Congo” as the Republic of the Congo, whose minister of posts, telecommunications and the digital economy, Frédéric Nze, took part. Belarus’s communications minister, Kirill Zalessky, and Cameroon’s posts and telecommunications minister, Minette Libom Li Likeng, signed for their governments; Lesotho and Mozambique were also reported as founding participants.
Ten of the 29 were African and four were in Latin America and the Caribbean—Brazil, Cuba, Nicaragua and Venezuela. The others were China, Russia, Kazakhstan, Kyrgyzstan, Tajikistan, Uzbekistan, Cambodia, Indonesia, Laos, Malaysia, Myanmar, Oman, Pakistan, Belarus and Serbia. Kazakhstan was reported as the only state known to have joined both this agreement and the U.S.-led Pax Silica initiative, in a July 2026 Observer Research Foundation compilation and in Radio Free Europe/Radio Liberty’s September account.
Later signatories and observers
The text stayed open for signature until 31 July 2026, so governments that signed after the Shanghai ceremony but inside that window are founding members under the agreement. Accession is only for states that join after the window closes.
On 30 July, Dominica’s ambassador signed in Beijing. On 31 July, Assistant Foreign Minister Liu Bin met representatives of Brunei, Togo, Iran, Sudan and Vietnam and witnessed their signatures; the ambassadors of Georgia and Tanzania signed the same day. With the original 29, that made 37 signatories in early-August counts. SESEC classed them as 16 in Asia, counting Russia and adding Iran, Vietnam and Brunei; 13 in Africa, adding Sudan, Tanzania and Togo; five in Latin America and the Caribbean, adding Dominica; and three in Europe—Belarus, Serbia and Georgia.
On 1 August 2026, Bangladesh’s foreign ministry said the government had decided to take part as an observer. The agreement allows observer status for non-member states and intergovernmental organizations only with Council approval; observers may attend plenaries on the chair’s invitation, at their own expense, and cannot vote. The ministry statement did not say the Council had already granted the status. On 9 September, Singapore’s Minister for Digital Development and Information, Josephine Teo, told Parliament that Singapore had been invited and would consider it alongside other engagements, including Pax Silica, which it had joined in December 2025. She said there was no ASEAN position on either arrangement and announced no decision.
Non-participation
The published rosters do not include the United States, any member of the G7, or any member state of the European Union. In Asia they also omit India, Japan, South Korea and Thailand. Thailand’s prime minister, Anutin Charnvirakul, attended the World Artificial Intelligence Conference that opened in Shanghai the day after the 16 July ceremony, but Thailand is not on the signatory lists.
The African signatories run from Algeria, Senegal, Cameroon and the Republic of the Congo to Ethiopia, Kenya, Lesotho, Mozambique, South Africa and Zambia, plus Sudan, Tanzania and Togo from 31 July. They do not include Nigeria, Egypt, Morocco or Ghana. In Latin America and the Caribbean the named signatories are Brazil, Cuba, Nicaragua, Venezuela and Dominica, not Mexico, Argentina, Chile or Colombia. Inside ASEAN, Cambodia, Indonesia, Laos, Malaysia and Myanmar signed on 16 July and Brunei and Vietnam on 31 July; the Philippines was not listed, and Singapore had not decided as of 9 September.
Legal framework
Purposes, principles and functions
The Agreement on the Establishment of the World Artificial Intelligence Cooperation Organization is the organization's constitutive treaty. Its preamble treats artificial intelligence as a driver of economic and social development that can accelerate the United Nations 2030 Agenda for Sustainable Development, calls for international public goods and cooperation so that AI develops in a "beneficial, safe and fair" direction, and records a desire to contribute to the purposes and principles of the Charter of the United Nations.
Article 2 states four purposes: a human-centric approach to the innovation, development, deployment and application of AI "for good and for all"; an open, fair, just and non-discriminatory environment for AI development, "with a view to unleashing the AI dividend"; bridging what the text calls the AI divide by making technologies and services more accessible; and collective mitigation of risks, challenges and impacts, alongside safe, secure, ethical and trustworthy AI.
Article 4 requires the organization to follow the purposes and principles of the UN Charter, respect the sovereignty of states and the diversity of civilizations, uphold equality, practice multilateralism, and harness the benefits of AI through extensive consultation and joint contribution.
Article 3 limits the functions exclusively to the civilian domain of AI. Within that limit the organization is to match supply and demand among member states; cooperate on capacity-building; align development strategies, governance rules and technical standards, facilitate dialogue, build policy consensus, promote interoperability and share best practices; strengthen science and technology cooperation and encourage an open-source ecosystem; support global AI governance and convergence with processes inside the UN framework; and communicate with other international organizations on possible cooperation. Konstantinos D. Magliveras, professor of public international law at the University of the Aegean, notes that the civilian-only limit is stated in the functions clause rather than among the purposes, and that several functions restate those purposes.
Legal status and depositary
Article 1 establishes the organization as an independent intergovernmental international organization, confers international legal personality and full legal capacity, and provides that it shall function in accordance with the agreement. Every contracting state is a member state. Article 5 creates two organs, a Council and a Secretariat, places the headquarters in the city of Shanghai, and allows subsidiary bodies and working groups. The Chinese foreign ministry likewise describes an independent intergovernmental organization headquartered in Shanghai.
The Government of the People's Republic of China is the depositary. It must notify all member states and the Secretary General of the date of entry into force and of signatures, ratifications, acceptances, approvals and accessions, and send certified copies to the governments of the other member states. The agreement is written in Arabic, Chinese, English, French, Russian and Spanish, all six texts being equally authentic. The English text cited here is the one published by the Geneva Internet Platform's Digital Watch Observatory; its articles correspond to Magliveras's reading of the agreement.
Entry into force, amendment and withdrawal
Article 14 opened the agreement for signature from 1 May 2026 until 31 July 2026. Signatory states must ratify, accept or approve it. After signature closed, the agreement remained open for accession. States that sign during the window are founding members. Any other state must apply to the Secretariat and may accede only with the Council's approval. Representatives of 29 countries signed on 16 July 2026; the Chinese foreign ministry stated that those countries became founding members.
The agreement enters into force on the date of deposit of the third instrument of ratification, acceptance or approval. For a state that ratifies, accepts, approves or accedes after that deposit, it enters into force on the date of that state's own deposit. The opening threshold is therefore tied to ratification, acceptance or approval, not to accession. Magliveras argues that three ratifications would allow a start with only three member states and that, under the quorum and majority rules, one favourable vote could then carry a decision, a result he calls difficult to reconcile with a multilateral organization.
Any member state may propose an amendment by sending it to the Secretary General at least two months before a Council plenary, for circulation. The plenary may adopt an amendment only by consensus; the two-thirds majority that Article 7 allows when consensus fails does not apply to amendments. An adopted amendment still requires ratification, acceptance or approval by every member state and enters into force only on deposit of the last such instrument, when it replaces the original provision.
Withdrawal is by written notification to the depositary and takes effect one year after receipt, unless the notification specifies a later date. It does not affect obligations that arose before it takes effect. Disputes on interpretation or application are to be settled by consultation among member states. The agreement does not confer jurisdiction on a court or an arbitral tribunal.
Institutional structure
The Agreement on the Establishment of the World Artificial Intelligence Cooperation Organization provides for two principal organs, a Council and a Secretariat, and places the headquarters in Shanghai, China. It also allows subsidiary bodies and working groups where they are needed to carry out the organization's functions. On 17 July 2026, the day after the signing ceremony, representatives of 26 countries attended what Indonesia's government described as an informal preparatory session in Shanghai; that meeting was not a plenary of the Council.
Council
Every member state sits on the Council, which the agreement makes the decision-making body. The Council sets overall strategy and regulations; reviews and approves the budget and final accounts, annual work plans and annual work reports; admits new member states and observers; appoints the secretary-general and the deputy secretaries-general; and establishes subsidiary bodies and working groups. It may also decide other matters that member states put forward, and exercise such further powers as the members determine are necessary to implement the agreement. It holds a plenary session once a year and elects a chairperson for a one-year term, with due regard to equitable geographical distribution. Each member state may designate one official coordinator to keep liaison with the Secretariat and to develop recommendations for the Council; meetings of the coordinators are convened under the Council's direction.
The Council adopts its own rules of procedure. Decisions require the presence of at least two-thirds of all member states and are taken by consensus to the furthest extent possible. If consensus cannot be reached after every effort has been exhausted, a decision may be made by a two-thirds majority of the member states present and voting, with each member state holding one vote. That majority fallback does not apply to amendments, which the plenary adopts by consensus.
Secretariat
The Secretariat is the permanent administrative body and works under the secretary-general, the chief administrative officer. It implements Council decisions; drafts the budget and final accounts, annual work plans and annual work reports; manages projects; coordinates with other international bodies; and handles other matters the Council assigns. The secretary-general is appointed by the Council from among nationals of member states, on nomination by member states, for a term of three years and may serve one additional term. Two deputy secretaries-general, also nationals of member states, are recommended by the secretary-general and appointed by the Council for the same term, three years with the possibility of one additional term. Other staff are appointed by the secretary-general under the organization's regulations. The paramount consideration is securing the highest standards of efficiency, competence and integrity, with due regard to recruitment on as wide a geographical basis as possible.
The offices were not filled at the launch. On 17 July 2026 the Chinese Foreign Ministry said China would work with the founding members for the early operation of the organization. The next day, coordinating minister for economic affairs Airlangga Hartarto proposed to Chinese foreign minister Wang Yi that Indonesia be considered for a senior secretariat post. On 7 August the Seconded European Standardization Expert in China (SESEC) project reported that leadership and governance were still being finalized. A 25 August analysis by the Pupin Initiative found staffing and funding undisclosed, no budget or contribution schedule public, and the timing of entry into force and domestic ratification still unclear. As of early October 2026, no secretary-general or deputy secretary-general had been publicly named, and the body was not yet fully operational.
Financing, privileges and disputes
Financial matters are governed by the agreement and by financial regulations or rules adopted by the Council. Funding is to come from annual member contributions, from voluntary contributions by governments, international organizations, corporations, individuals and other entities, and from lawful revenue of cooperative projects. Contributions may be accepted only with the Council's permission, and not if they would prejudice, restrict, distort or alter the organization's purposes, objectives or functions. The Council sets the annual scale of assessment with reference to members' contribution classes in the Universal Postal Union, and it may also take account of economic development and capacity to pay. The Union, a specialized agency of the United Nations, finances its own expenses through those classes: a joining country chooses among ten classes from one to 50 units, with a further half-unit class for least developed countries. The agreement uses that system as a reference and does not itself fix unit values or assign classes. As of 25 August 2026, no financial regulations, budget or contribution schedule had been published.
The organization, its staff and representatives of member states are to enjoy, in every member state, the privileges and immunities necessary for its functions and goals, with the extent left to separate agreements. Any dispute over the interpretation or application of the agreement is to be settled through consultation among member states.
Cooperation activities
By early October 2026, cooperation linked to the World Artificial Intelligence Cooperation Organization consisted of functions in the founding agreement and separate pledges by the government of China. The organization had not announced a secretary-general or a project of its own. On 17 July, Foreign Ministry spokesperson Lin Jian said China would work with the founding members toward early operation.
Capacity-building
The agreement seeks a human-centric approach to artificial intelligence "for good and for all," an open, fair, just and non-discriminatory environment, a narrower AI divide, and safe, secure, ethical and trustworthy systems. Its functions, limited to civilian AI, include a platform matching supply and demand among member states and international cooperation on capacity-building. Project management sits with a secretariat under a secretary-general appointed by the Council. Officials quoted by Xinhua said future work would treat capacity-building as a priority, through policy exchange, technical cooperation, training and joint research.
The numbers given in Shanghai were Chinese commitments, and they are not in the agreement. On 17 July, President Xi Jinping said that over five years China would offer developing countries 5,000 places in AI training and seminar programmes; develop international AI application cooperation centres with the Association of Southeast Asian Nations, the League of Arab States, the African Union, the Community of Latin American and Caribbean States, the Shanghai Cooperation Organisation and BRICS; and let 30 countries use MAZU, a Chinese AI-powered meteorological warning system. The offers continued an earlier track. On 1 July 2024 the United Nations General Assembly adopted resolution 78/311 on AI capacity-building, limited to non-military uses and noting benefits and risks of open-source software, open models and open data. In September 2024 China issued an AI Capacity-Building Action Plan for Good and for All, offering joint laboratories, cooperation on models and applications, a cooperation platform and workshops for developing countries, in support of that resolution.
Technical cooperation and open-source ecosystems
The agreement also calls for alignment of strategies, governance rules and technical standards, interoperability and shared practices, and for science and technology cooperation that encourages an open-source ecosystem. It does not define open source, require release of weights or training data, or name any firm. Xi urged governments to encourage open source, openness, collaboration and sharing. The Centre for International Governance Innovation noted that the speech, the conference chair's statement and the announcement of the organization did not explicitly commit China, or the organization, to open-source or open-weight models.
That debate is about the industry, not an organizational programme. A 2025 primer from the OECD and the Global Partnership on Artificial Intelligence described openness as a spectrum across code, data and weights, and reported that more than half of commercially available foundation models were open-weight, led by developers in the United States, China and France, including DeepSeek's R1, Alibaba Group's Qwen series and OpenAI's GPT-OSS. A Stanford Institute for Human-Centered Artificial Intelligence brief found that many Chinese open-weight systems had first been built on Meta Platforms' Llama models, before DeepSeek, Alibaba and later Baidu distributed weights of their own. On the July 2026 meetings, CIGI reported that Moonshot AI presented Kimi K3 as remaining open-weight and that DeepSeek's founder had indicated further open releases. Both were still described as behind the strongest models associated with OpenAI and Anthropic. A Brookings discussion described open-weight work as dominated by Chinese firms including Moonshot, DeepSeek and Z.ai, set against paid access to systems from OpenAI and Anthropic. None of those releases was identified as an activity of the organization.
Engagement with other organizations
The agreement mandates support for convergence with processes inside the United Nations framework and communication with other international organizations, and it requires the organization to follow the Charter of the United Nations. António Guterres, the UN secretary-general, attended the 16 July signing in Shanghai. He addressed the conference the next day, and Lin Jian said China would keep working on governance under the UN, including resolution 78/311, the Global Digital Compact and the Sustainable Development Goals. Xinhua reported that members would later explore cooperation with the UN and other international and regional bodies, and that joining would not affect third-party cooperation or add unnecessary constraints on national AI policy.
The centres Xi named are Chinese projects with regional groups, not organs of the organization. Some were already operating. In May 2026, before the signing, Arindrajit Basu of the Carnegie Endowment for International Peace described a China–BRICS AI development and cooperation centre opened in Shanghai in January 2025 and a China–Laos AI innovation cooperation centre, and wrote that Beijing would likely use the proposed body to enlarge that network. CIGI contrasted the agreement's emphasis on capacity-building, openness and the UN with smaller arrangements such as the Global Partnership on AI, as a difference of design rather than a relationship the organization had started.
Place in global AI governance
United Nations processes
The agreement establishes the organization as an independent intergovernmental body, not a specialized agency of the United Nations. It is to follow the purposes and principles of the Charter of the United Nations, and one function is “to enhance global AI governance, provide support for and promote convergence with the process within the U.N. framework,” while keeping contact with other international organizations. The Center for Strategic and International Studies describes it as operating outside the UN system.
António Guterres was in Shanghai for the July 2026 World Artificial Intelligence Conference. China’s foreign ministry said he attended and witnessed the 16 July signing; the United Nations was not a party. The next day he called it the “natural development” of the Global AI Governance Initiative that Xi Jinping launched in 2023, and invoked the Global Digital Compact’s line that technology shaping humanity “must be shaped by all humanity.”
The compact, adopted with the Pact for the Future by the General Assembly on 22 September 2024 (A/RES/79/1) at the Summit of the Future, is the UN framework for digital cooperation and for governing artificial intelligence. It committed states to a scientific panel and a dialogue, following the September 2024 report of the secretary-general’s High-level Advisory Body, Governing AI for Humanity, which recommended light complementary mechanisms. Resolution A/RES/79/325, adopted by consensus on 26 August 2025, created both. The Independent International Scientific Panel on AI has 40 members and assesses the non-military domain; the UN states that it will not set rules, enforce standards or prescribe policy. Yoshua Bengio and Maria Ressa were elected co-chairs on 3 March 2026; a preliminary report followed on 1 July. The Global Dialogue on AI Governance is not a negotiating forum and closes with a co-chairs’ summary rather than a binding text. Its first session, in Geneva on 6–7 July 2026, drew more than 3,000 participants from 163 countries; the next is in New York City on 3–4 May 2027. China’s action plan of 26 July 2025 asked that the UN be the main channel and that both mechanisms start promptly, and CSIS describes a dual track of support for those bodies and for this organization outside them.
Other intergovernmental initiatives
Parallel efforts differ in membership and legal force. The OECD AI Principles, adopted in May 2019 and updated in May 2024, are a Council recommendation. The OECD says its definition of an AI system is used by the European Union, the Council of Europe, the United States and the United Nations, and presents the principles as a basis for interoperability. In New Delhi on 3–4 July 2024, ministers of the Global Partnership on Artificial Intelligence reaffirmed those principles and the UNESCO Recommendation on the Ethics of AI, and agreed an integrated partnership with the OECD on an equal footing for members. The G7 Hiroshima AI process, begun in 2023, is voluntary; CSIS reports endorsement by 49 countries and a 2025 OECD-run reporting framework for developers’ risk-management practices.
The Council of Europe Framework Convention on Artificial Intelligence, opened for signature on 5 September 2024, is described by the Council as the first legally binding international treaty in the field. It enters into force only after five signatories, including at least three member states, consent to be bound; the Council’s page lists the European Union as a party. The EU Artificial Intelligence Act, Regulation (EU) 2024/1689 of 13 June 2024, harmonises rules for the internal market. It entered into force on the twentieth day after publication on 12 July 2024, with duties phased in from 2025.
Asia-Pacific Economic Cooperation leaders endorsed an AI Initiative for 2026–2030 in the 2025 Gyeongju Declaration, covering innovation, capacity-building and infrastructure through APEC fora rather than a treaty.-initiative-(2026-2030)) The Bletchley Declaration of 1–2 November 2023, from the AI Safety Summit, committed governments to cooperate on frontier-AI risks; CSIS describes later meetings in Seoul (2024), Paris (2025) and New Delhi (2026) as shifting that series from safety toward access and deployment. Premier Li Qiang proposed the body in July 2025 with the thirteen-point action plan. Before the signing, Arindrajit Basu of the Carnegie Endowment for International Peace argued that Beijing was moving from infrastructure and standards toward institutions that could shape rules and align positions at the United Nations. CSIS sets the signing beside the Geneva dialogue and a U.S. Pax Silica supply-chain summit as three separate designs; Pax Silica is covered under strategic competition.
Standards and interoperability
The agreement does not adopt technical standards or duties for companies. It assigns the organization “alignment and coordination on AI development strategies, governance rules and technical standards,” dialogue, policy consensus, interoperability and shared practices, and an open-source ecosystem, limited to civilian AI. China’s 2025 action plan, which is not a decision of the organization, separately supported dialogue among standards bodies and a role for the International Telecommunication Union, the International Organization for Standardization and the International Electrotechnical Commission.
Analysts treat that as cooperation and capacity-building, not binding regulation. In December 2025 Elizabeth Gibney wrote in Nature that governments had not agreed AI regulations and that, for many countries, binding rules were still being developed, even as China sought an international system to govern the technology’s use. CSIS reported that the organization had adopted no binding commitments and published no standards, while describing the Hiroshima process as voluntary and the UN dialogue as issuing no binding recommendations.
Strategic competition
Governments and analysts have placed the organization in competition between the United States and China over artificial intelligence, its supply chain and access to models. In The Diplomat, Catherine Putz called Pax Silica and the organization competing alignments—one concerned with physical supply chains, the other with governance and norms—and wrote that a "new cold war" framing depended on whether governments were forced to choose. The Straits Times described the organization as seen as a counterpoint to the U.S. initiative. The U.S. State Department has instead called Pax Silica cooperation among partners "who want to remain competitive," and has said it "is not about isolating others."
Pax Silica and allied supply-chain initiatives
Pax Silica is a State Department initiative opened at a Washington summit on 12 December 2025 by Under Secretary Jacob Helberg. The department described a silicon supply chain "from critical minerals and energy inputs to advanced manufacturing, semiconductors, AI infrastructure, and logistics," and called the effort "a new kind of international grouping and partnership," not a treaty organization. Counterparts came from Japan, South Korea, Singapore, the Netherlands, Israel, the United Arab Emirates, the United Kingdom and Australia, with guests including the European Union and Taiwan. The South China Morning Post reported that Singapore was one of seven signatories at the launch.
The declaration, which does not name China, pledges signatories to reduce "excessive dependencies," address non-market practices, and protect sensitive technologies from "undue access, influence, or control." Putz wrote that the language on dependencies and "reliable partners" was aimed at Beijing. Reuters called the framework non-binding and reported that by August 2026 about two dozen countries had joined, including Japan, Australia, South Korea and Kazakhstan, which Washington had highlighted for its critical minerals. Reuters said the arrangement was meant to move allies toward joint projects and export controls, reduce reliance on adversaries for minerals, models and the semiconductor chips that run them, and open shared investment in AI-related projects.
Export controls and access to models
U.S. limits on advanced chips are the setting for Chinese statements that AI should not be closed to most countries. On 17 October 2023 the Commerce Department's Bureau of Industry and Security updated rules issued on 7 October 2022, restricting exports to China of advanced computing semiconductors, semiconductor manufacturing equipment and supercomputing items, citing military uses of advanced computing and artificial intelligence. On 26 July 2025 Premier Li Qiang, without naming the United States, said "technological monopoly, controls and restrictions" would make AI "an exclusive game for a small number of countries and enterprises," and proposed a global cooperation body. CNN placed the remarks against the Trump administration's AI action plan and U.S. limits on chips, including an Nvidia processor.
On 17 July 2026, at the World Artificial Intelligence Conference in Shanghai, President Xi Jinping said development "should not be a solo performance by a single country, but a symphony of international cooperation," urged open source and sharing, and called on countries to "jointly oppose overstretching the national security concept in the field of AI." Reuters later connected model access to the rivalry, writing that Chinese open-weight systems had gained on proprietary models from OpenAI and Anthropic, and that Beijing was weighing limits on overseas access to some leading Chinese models. CNN's account of the 2025 conference reported that a model from DeepSeek had outpaced systems from Meta Platforms and Anthropic.
Pressure on prospective members
On 14 August 2026 Reuters reported an undated draft State Department letter to the 35 signatories of a June U.S. "AI Opportunity Statement," including Pax Silica participants and other governments seeking AI cooperation with Washington. Reuters read the draft as a warning that joining Beijing's framework would mean exclusion from the U.S.-led coalition. The quoted text did not name this organization or China: "To be part of everything is to be part of nothing. Signature of the Pax Silica Declaration is not merely a membership subscription, but a commitment," and that commitment "cannot be held alongside membership in duplicative initiatives whose expectations conflict with our own." Reuters could not learn whether the letter would be sent or amended, and the department declined to comment on "purportedly leaked internal documents." An unnamed U.S. official said the draft was to show that "you can't have it both ways," and that it was hard to be a trusted partner while joining "an initiative designed by China to advance a competing vision for AI"—an account of a draft, not a step Reuters found had been taken.
Reuters identified Kazakhstan as the only country then known to be in both; its embassy did not respond. China's embassy said Beijing opposed "politicizing trade and technology issues," because that would "stifle global AI advances and serve no one's interests." As of 8 September 2026, Putz reported that the letter still did not appear to have been sent, and that Uzbekistan's ambassador, Furqat Sidikov, said Tashkent had asked to join Pax Silica and called "red lines" on which forums a country may join "counterproductive."
On 9 September 2026, Minister for Digital Development and Information Josephine Teo told Parliament that Singapore had been invited, that ASEAN had no position on either initiative, and that the invitation would be considered with Singapore's other partnerships. She said one-to-one comparisons were "less meaningful" than those relationships. Analysts quoted by the South China Morning Post said membership could enlarge Singapore's voice in AI governance but might draw an unfriendly U.S. response and affect its Pax Silica role. Chong Ja Ian, of the National University of Singapore, said resulting U.S. restrictions might outweigh the benefits; William Heidlage of BowerGroupAsia called the two efforts complementary for Singapore.
Reception
Policy analysis
Before the July 2026 signing, commentary asked whether a China-proposed body would regulate artificial intelligence or mainly organize cooperation. In a 1 December 2025 Nature explainer, Elizabeth Gibney reported that the plan was not expected to impose directly enforceable rules. She described a forum for gradual alignment, weighted toward technical cooperation and capacity-building, and reported the proposal's pledge to proceed while "fully respecting the differences in national policies and practices." A Nature editorial on 10 December 2025 said dialogues had not yielded globally agreed rules, that creating the organization was in all countries' interests, and that governments should engage rather than treat AI as a race for dominance. It cited the International Atomic Energy Agency model of agreed limits plus inspection.
In May 2026, Arindrajit Basu wrote for the Carnegie Endowment for International Peace that Beijing was likely to use the body to expand bilateral AI centres and to assemble a consensus for later coordination at the United Nations. He treated that as a move from infrastructure exports toward state-centric norms, and warned that large language models from Chinese firms could censor content tied to liberal political claims wherever they were deployed. Governments in the Global South, he wrote, had not endorsed the architecture wholesale.
After the signing, Christopher Collins and Aalok Mehta of the Center for Strategic and International Studies placed the organization beside Pax Silica and a UN dialogue, and argued that the dispute was as much about who would write rules as about their content. No G7 economy or European Union member joined. They tied that membership to a 2024 UN advisory finding that 118 countries, mostly in the Global South, sat outside the main recent AI initiatives. Possible paths, they wrote, resembled the Belt and Road Initiative or the Asian Infrastructure Investment Bank; whether any OECD member joined would be an early test. In The Diplomat, Gerald Mako called the launch a turn toward bloc-based governance and wrote that the absence of Western or strongly Western-aligned founders pointed to parallel ecosystems for standards, data and infrastructure. Michael Caster, also in The Diplomat, called it China's most ambitious bid to formalize influence over deployment: training, standards work and open-source promotion could normalize a sovereignty-centred approach, and spread Chinese models, without replacing other forums. He described the body as still aspirational and said domestic digital controls warranted caution.
Commentary in member states
Officials in Indonesia presented membership as a way to narrow an AI divide, not as a choice against the United States. Deputy Minister Angga Raka Prabowo said governance should become financing, technology transfer and capacity-building for small firms and local languages. Foreign Ministry spokesperson Vahd Nabyl Achmad Mulachela told Kompas that developing countries could use the body to shape policy and standards, while Pax Silica concerned supply chains, and that without computing access a state risked remaining a "rule-taker." Analyst Christian Guntur Lebang said Indonesia lagged major economies across energy, chips, cloud, models and use.
Singapore had not joined. On 9 September 2026, Minister for Digital Development and Information Josephine Teo told Parliament that Singapore had an invitation, that ASEAN had no position on the organization or Pax Silica, and that any decision would turn on objectives, overlap with existing work and national interests. The Diplomat described Kazakhstan, then reported as the only state in both Pax Silica and the new organization, as a multivector hedge that a U.S. draft warning against "duplicative initiatives" could strain. In Serbia, the Pupin Initiative argued in August 2026 that founding membership was legally thin and politically costly for an EU candidate whose compute, exports and AI strategy were tied to Western partners, and that Belgrade had identified neither a signatory nor a concrete deliverable.
Criticisms and unresolved questions
Collins and Mehta wrote in September 2026 that the organization had adopted no binding commitments and published no standards, and that secretariat arrangements, budget, decision rules, enforcement and a technical work programme remained sparse. In their account, Chinese offers of training places and regional centres were government pledges beside the launch, not yet the organization's programmes. They wrote that a secretariat, non-Chinese funding, shared leadership and verifiable projects were still required, though an annual forum could shape rules if rival offers stayed limited to supply chains. Konstantinos D. Magliveras argued that entry into force on the third ratification could let three members, and failing consensus a single vote, carry early decisions. Basu's warning on content controls in Chinese models, Caster's reading of open-source promotion as a standards channel, and Mako's forecast of parallel ecosystems were claims about influence, not findings that the agreement imposed content rules. The Nature editorial and the CSIS account of states left out of safety-focused clubs treated a development-oriented forum as filling that gap, if engagement produced shared leadership and projects members could check.
By the Factrail desk · Method: how we source and score · Last reviewed
Related people
Related organizations
- United Nationsnamed 20 times
- Pax Silicanamed 17 times
- Carnegie Endowment for International Peacenamed 4 times
- The Centrenamed 3 times
- South China Morning Postnamed 2 times
- Shanghai Cooperation Organisationnamed 2 times
- Community of Latin American and Caribbean Statesnamed 2 times
- The Straits Timesnamed 2 times
- Policy Exchange
- Ministry of Foreign Affairs
- The Independent
Linked pages by kind
Each bar counts the pages of that kind listed in the blocks here (a full list is shown as "N+"). Counts, not a rating.
- People8
- Organizations11