GPAI Global AI Governance Overview

Overview

The Global Partnership on Artificial Intelligence (GPAI) was officially launched in June 2020, representing a landmark achievement in international digital diplomacy. Conceived through the joint leadership of Canada and France and born out of discussions within the G7, GPAI was established to provide a permanent platform where experts from science, industry, civil society, and government could collaborate to promote 'responsible AI'. This concept of responsible AI is defined as technology that is human-centric, respects fundamental human rights, and remains consistent with democratic values. GPAI's unique value proposition lies in its multi-stakeholder nature, which distinguishes it from purely intergovernmental organizations. By bringing together leading experts to work on practical, applied projects, GPAI aims to inform policy through evidence-based research. As of 2024, GPAI has entered a new era by integrating its governance and secretariat functions with the Organisation for Economic Co-operation and Development (OECD). This 'Integrated Partnership' was designed to create a streamlined global hub for AI, allowing for a more rapid response to the fast-paced evolution of AI technologies, such as generative models and large-scale foundation systems, while ensuring that international safety standards and ethical oversight remain at the forefront of global development.

AI Governance Mandate

GPAI's mandate is fundamentally rooted in the OECD Principles on Artificial Intelligence, which serve as the international gold standard for trustworthy AI. The partnership's primary objective is to support the development of AI that respects human rights and fosters public trust across all sectors of society. This mandate is operationalized through four key pillars: (1) Responsible AI, which focuses on ethics and safety; (2) Data Governance, which explores the legal and technical frameworks for sharing data; (3) Future of Work, which examines the socio-economic impacts of automation; and (4) Innovation and Commercialization, which seeks to lower barriers for small and medium-sized enterprises (SMEs). By facilitating international cooperation, GPAI seeks to ensure that AI technologies are developed and deployed in a manner that benefits all citizens, reduces global inequalities, and addresses existential challenges such as climate change and public health crises. The governance structure is specifically designed to ensure that technical expertise from the scientific community directly informs the political commitments and legislative agendas of member nations, creating a feedback loop between innovation and regulation.

Key AI Instruments

GPAI operates through several high-level instruments and declarations that set the strategic direction for its members and influence the broader international community. The New Delhi Ministerial Declaration (2023) marked a pivotal moment in the partnership's history, where members committed to addressing the specific risks posed by generative AI, including misinformation and the potential for algorithmic bias. This declaration also emphasized the need for 'sovereign AI' capabilities and equitable access to AI resources for the Global South, ensuring that the benefits of AI are not concentrated in a few geographic regions. This followed the Tokyo Ministerial Declaration (2022), which focused on the concept of 'Trustworthy AI' and the necessity of creating interoperable governance frameworks that allow for the cross-border flow of data with trust. While these declarations are non-binding, they carry significant political weight and serve to align national AI strategies with international norms. Beyond these declarations, GPAI produces a steady stream of technical reports and frameworks, such as the 'Data Justice' report and the 'Responsible AI Strategy for the Environment,' which provide actionable guidance for implementing ethical AI principles in specific industrial and social sectors.

Implementation Mechanisms

The practical implementation of GPAI's goals is driven by its specialized Working Groups and its network of Expert Support Centers. The Working Groups are the engine room of the partnership, composed of hundreds of experts from diverse backgrounds who conduct rigorous research and develop pilot projects. These groups include: (1) The Responsible AI Working Group, which focuses on AI safety, ethics, and the protection of fundamental rights; (2) The Data Governance Working Group, which investigates mechanisms for secure and equitable data sharing, including the development of data trusts and data commons; (3) The Future of Work Working Group, which analyzes the impact of AI on labor markets and identifies strategies for workforce upskilling and social protection; and (4) The Innovation and Commercialization Working Group, which supports the transition of AI research into sustainable business models. Supporting these groups are the Montreal Centre of Excellence (CEIMIA) and the Paris Centre (INRIA), which provide the administrative, technical, and financial support necessary to execute complex projects. These centers ensure that GPAI’s theoretical work results in tangible tools, such as safety evaluation benchmarks and transparency documentation templates, that can be adopted by developers and regulators worldwide.

Member Engagement

GPAI's membership has expanded significantly since its inception, now including 29 members, including the European Union and major economies across the Americas, Europe, and the Asia-Pacific region. A core priority of GPAI's engagement strategy is the active inclusion of the 'Global South.' By recruiting and involving developing nations, GPAI aims to prevent a 'digital divide' and ensure that AI governance is not a Western-centric endeavor. Members engage through the annual GPAI Summit, a high-profile event where the Council and Steering Committee make critical decisions on the partnership’s strategic priorities for the coming year. The integration with the OECD in 2024 has further expanded this engagement capacity, allowing GPAI to tap into the OECD's broader diplomatic network and its deep policy expertise. This synergy increases the impact of GPAI's expert-led initiatives, as they are now more closely linked to the OECD's policy recommendations which are frequently adopted into national law by member states. This collaborative environment fosters a 'community of practice' where nations can share successes and failures in AI regulation, leading to more harmonized global standards.

Current Focus Areas

In the current technological landscape, GPAI has prioritized the governance of Generative AI and Foundation Models. Recognizing the unprecedented speed at which these technologies are being adopted, GPAI is working on the development of robust safety testing protocols and stringent transparency requirements to mitigate risks such as the mass production of deepfakes and the reinforcement of societal biases. Another major focus is 'Data Justice,' a framework that seeks to ensure that the benefits of AI-driven data analysis are shared fairly and that marginalized communities are protected from exploitative data practices. Furthermore, GPAI is increasingly focusing on the intersection of AI and Climate Change. This involves developing frameworks to measure and mitigate the environmental footprint of training large-scale AI models, while simultaneously promoting the use of AI for environmental monitoring, biodiversity preservation, and the optimization of renewable energy grids. These focus areas reflect GPAI's commitment to addressing both the immediate risks and the long-term societal opportunities presented by the AI revolution.

Future Directions

The future of GPAI is defined by its new status as an integrated partner of the OECD, a move finalized in 2024 to create a more agile and efficient organization. This integration is intended to position GPAI as the primary global laboratory for testing new governance ideas, such as 'regulatory sandboxes' for international AI collaboration. Future activities will likely focus on the operationalization of the 'Hiroshima AI Process' and the development of international benchmarks for AI safety evaluation that can be used by both governments and private companies. GPAI will continue to advocate for a unified approach to AI governance that prioritizes human dignity and the common good, playing a key role in the United Nations' Global Digital Compact. As AI systems become more autonomous, GPAI is expected to lead the discourse on AI liability and the legal status of AI-generated content. The partnership remains dedicated to ensuring that the voice of the scientific and technical community remains central to the global AI policy discourse, preventing political interests from overshadowing technical realities and ethical imperatives.

Resources

Standards Tracked

All 2 instruments currently tracked for GPAI.