THE launch of the World Artificial Intelligence Cooperation Organisation in Shanghai on July 16, 2026, may mark an important moment in global AI governance. Initiated by China, the body aims to promote cooperation and keep AI beneficial, safe and fair. It has attracted 29 countries, mostly from the Global South, and is expected to grow. To become globally effective, WAICO must broaden its membership across North and South and evolve into a genuinely multilateral institution. President Xi Jinping summarised its ambition: AI development, he said, should not be “a solo performance by a single country, but a symphony of international cooperation”.
WAICO emerges amid a technological dilemma. Like nuclear technology, AI promises major benefits but also raises fears of misuse, instability and harm. The new organisation’s task is not simply to control AI, but to also manage risks while widening access to benefits without discrimination. Its founders should therefore study the International Atomic Energy Agency (IAEA), which faced a similar challenge almost seven decades ago.
The atomic bomb demonstrated that nuclear technology could transform the strategic balance between nations and threaten civilisation; the same science also offered advances in medicine, agriculture, industry and energy. The challenge was to prevent destructive applications and share peaceful benefits. President Dwight Eisenhower’s ‘Atoms for Peace’ address to the UN General Assembly in 1953 proposed an international agency to make nuclear technology available for peaceful purposes “for the benefit of all mankind”. His initiative led to the IAEA’s creation in 1957.
The IAEA was never merely a nuclear policeman. Its statute covered peaceful applications, scientific exchange, training, safeguards and safety standards. Its founding principle was to make nuclear technology safer without denying humanity its benefits — a central lesson for WAICO. Safeguards became a core IAEA function, assuring states that civilian nuclear materials and activities were not diverted to military purposes. After the 1968 Nuclear Non-Proliferation Treaty, they became central to the global non-proliferation regime. Yet the Agency also helped countries acquire knowledge, train personnel, build infrastructure, and use nuclear science peacefully. An organisation that only prohibits will meet resistance; one that also advances legitimate development creates stronger cooperation.
The IAEA’s institutional logic offers five useful lessons for the World Artificial Intelligence Cooperation Organisation.
The IAEA developed complementary roles in verification, peaceful uses, technical cooperation, safety and security. As nuclear technology spread, countries needed not only non-proliferation controls but also secure materials, safe operations, capable regulators, trained personnel and strong national institutions.
AI can transform healthcare, education, agriculture, energy, manufacturing, public administration and research, helping developing countries overcome barriers. Yet it also brings job displacement, misinformation, cyberattacks, fraud and potentially harmful autonomous decision-making. The question is how to keep this power aligned with progress without letting its destructive potential escape control. While the IAEA experience is relevant, WAICO should not copy the Agency. AI is software-based, diffuse, fast-moving and developed by thousands of private actors. Nevertheless, the IAEA’s institutional logic offers five useful lessons.
Risk management cannot be separated from development: Governance cannot be framed only as restriction. For many developing countries, the priority is access. A regime in which advanced states possess AI while others merely follow rules designed elsewhere will deepen inequality. So, WAICO should make capacity building a central pillar. China’s plan to offer developing countries 5,000 AI training and seminar opportunities over five years and establish international application centres could become a permanent multilateral programme. It should build national capacity — universities, research institutions, regulators, data infrastructure, etc. — much as IAEA technical cooperation has done in the nuclear field.
Develop an AI safeguards concept: Trust cannot rest on declarations. AI safeguards should not block access but identify exceptionally powerful capabilities, manage their risks, and verify commitments. A risk-based system could grade capabilities, set standards, require transparency, establish testing methods, and independently assess declared advanced systems through continuous monitoring and peer review like the IAEA verification and peer review system. It should focus on consequential applications, not intrusively control every system.
AI safety should be an international public good: This is necessary because failures and malicious use can cross borders. WAICO could develop shared principles, standards, testing methods and incident-reporting mechanisms, and enable countries to share failures and near misses so others can prevent their recurrence. IAEA safety standards and operating experience systems show how consensus guidance and corrective learning can reduce risk without assigning blame.
Security must be distinguished from safety: Safety addresses accidents, failures and unintended harm; security addresses deliberate misuse, manipulation and attack. Distinguishing between them would support clearer legal, regulatory and ethical responses.
Standards aren’t enough — people and institutions matter: Standards work only when countries have capable institutions and people. WAICO could create regional centres of excellence along with training, fellowships, laboratories and regulatory capacity programmes, helping developing countries implement commitments similar to IAEA technical cooperation programmes.
WAICO should learn from the IAEA but not become the IAEA of AI. AI differs fundamentally from nuclear technology. Nuclear material can be physically controlled and is concentrated in capital-intensive facilities; AI software can be copied, modified, and distributed instantly. WAICO cannot transplant the IAEA model, but it can adopt its balanced approach: enable beneficial use, build capacity, set safety and security standards, and prevent misuse.
The international community has spent nearly 70 years learning to govern a beneficial but dangerous technology. The IAEA also has its limitations and is politically constrained, yet WAICO should study its successes, failures, and adaptations rather than reinvent the wheel.
Eisenhower sought to turn the atom from an instrument of destruction into a force for peace. Today, the challenge is to ensure that AI serves humanity rather than subordinates or harms it. WAICO should place safety and security alongside development and capacity building; promote cooperation without blocking developing country access; set international standards that respect national contexts; and focus oversight on the gravest risks rather than every application.
In 1953, the world answered Eisenhower’s call and created the IAEA. Today, Xi has called for AI to become “an important driver for shared prosperity and common security”. The world should respond with equal seriousness.
Reference Link:- https://www.dawn.com/news/2032521