Global Partnership on Artificial Intelligence (GPAI)

The Global Partnership on Artificial Intelligence (GPAI) originated in a joint proposal advanced by France and Canada at the 2018 G7 Charlevoix Summit, where leaders endorsed the idea of an international study group on AI modelled loosely on the Intergovernmental Panel on Climate Change. The concept matured through 2019 and was formally launched on 15 June 2020 by fifteen founding members, anchored on the OECD Recommendation on Artificial Intelligence (adopted 22 May 2019), the first intergovernmental standard on AI. GPAI's legal character is that of a voluntary, non-binding partnership rather than a treaty organisation; its members are sovereign states (and the European Union) committed to the OECD principles of inclusive growth, human-centred values, transparency, robustness, and accountability. The OECD in Paris hosts GPAI's dedicated secretariat, integrating the initiative into the broader OECD.AI Policy Observatory ecosystem while preserving its distinct membership and governance.

Procedurally, GPAI operates through a layered structure that separates political direction from expert work. A Council, composed of ministerial-level representatives of member states, sets strategic priorities and admits new members. A Steering Committee translates that direction into a work programme and oversees the experts. The substantive output is produced by multistakeholder Working Groups, whose members are drawn from government, industry, civil society, academia, and international organisations, and who serve in their individual expert capacity rather than as instructed delegates. Working Group deliverables—reports, pilot projects, and policy recommendations—are channelled through annual plenary summits and ministerial meetings, where the partnership endorses findings and shapes the following year's agenda. Decisions on membership and direction proceed by consensus, consistent with GPAI's non-binding ethos.

The Working Groups have been organised around thematic pillars: Responsible AI, Data Governance, the Future of Work, and Innovation and Commercialisation, with Responsible AI further encompassing dedicated streams on AI and pandemic response and on climate action. Two physical Centres of Expertise support this work—one in Montréal, operated by the International Centre of Expertise in Montréal for the Advancement of Artificial Intelligence (CEIMIA), and one in Paris, hosted by INRIA—providing administrative and research backbone to the expert groups. In November 2023, GPAI and the OECD announced an integrated partnership designed to bring GPAI members and the OECD's AI work under a more unified framework, broadening participation to developing and emerging economies and reducing duplication between the two Paris-based structures.

By 2024 GPAI's membership had grown to roughly forty-four states plus the European Union, spanning North America, Europe, the Indo-Pacific, Latin America, and Africa. India joined as a founding member in 2020 and assumed the rotating Council chair, hosting the GPAI Summit in New Delhi from 12 to 14 December 2023, where the New Delhi Declaration was adopted; France had hosted the inaugural summit in 2020, Japan in 2022. The United States, the United Kingdom, Germany, South Korea, Singapore, Brazil, and Mexico count among the participants, alongside the OECD Secretary-General's office. The New Delhi Summit advanced initiatives on collaborative AI for the Global South and a framework for governing generative AI, reflecting India's emphasis on equitable access and application-driven development.

GPAI must be distinguished from several adjacent instruments. It is not the OECD AI Principles themselves—rather, it is the multistakeholder vehicle that operationalises those principles through applied projects. It differs from the EU AI Act, a binding regional regulation entering into force in 2024, in that GPAI produces no enforceable law. It is separate from the UN High-Level Advisory Body on Artificial Intelligence and from UNESCO's 2021 Recommendation on the Ethics of AI, both of which sit within the United Nations system rather than the OECD orbit. It is likewise distinct from the AI Safety Summit process initiated at Bletchley Park in November 2023, which concentrates on frontier-model catastrophic risk, whereas GPAI's mandate is broader and more development-oriented.

GPAI has attracted critique on grounds familiar to multistakeholder governance. Its non-binding character means recommendations carry no compliance mechanism, leading some observers to question its practical leverage against the binding force of the EU AI Act or national legislation. The exclusion of the People's Republic of China—a consequence of the partnership's explicit grounding in democratic values and human rights—limits its claim to global coverage of AI capacity. The 2023 integration with the OECD prompted debate over whether GPAI would retain its distinct multistakeholder identity or be subsumed into an intergovernmental body where civil-society and academic voices weigh less heavily. The rapid emergence of generative AI after late 2022 also pressured GPAI to reorient work programmes faster than its annual summit cadence comfortably allowed.

For the working practitioner—the desk officer tracking technology diplomacy, the policy researcher mapping the AI governance landscape, or the UPSC aspirant preparing GS Paper II on international institutions—GPAI is a reference point for understanding how like-minded democracies coordinate on emerging technology without ceding sovereignty to a treaty regime. It exemplifies the soft-law, principles-first model that increasingly characterises global digital governance, sitting between purely normative declarations and binding regulation. India's chairmanship signals the strategic weight emerging economies now command in shaping AI norms, and GPAI's evolving relationship with the OECD, the G7 Hiroshima AI Process, and the UN system offers a live case study in institutional layering and the contest to set the rules of the AI era.