GPAI AI Regulation Overview

Overview

The Global Partnership on Artificial Intelligence (GPAI) is a landmark international multi-stakeholder initiative launched in June 2020 to bridge the gap between high-level policy principles and the practical implementation of artificial intelligence. Conceived during the 2018 and 2019 G7 Summits under the leadership of Canada and France, GPAI serves as a collaborative platform for member nations and invited experts to support responsible, human-centric AI development. Grounded in the OECD Recommendation of the Council on Artificial Intelligence, the initiative emphasizes human rights, inclusion, diversity, innovation, and economic growth, fostering international cooperation on complex AI challenges that individual nations cannot resolve alone.

Founded by fifteen initial members, GPAI's membership has expanded to 44 countries. In July 2024, GPAI established an integrated partnership with the OECD, housing a permanent secretariat at OECD headquarters in Paris. In 2025, GPAI expanded its policy focus to address emerging technological frontiers through the GPAI Agentic AI Systems Framework initiative led by the Tokyo Expert Support Center, ensuring that autonomous AI systems with independent decision-making capabilities remain aligned with human values, transparency, and safety.

Regulatory Approach

GPAI adopts a soft law, multi-stakeholder, and project-based approach to AI governance rather than enacting legally binding statutory mandates. Its governance framework is anchored in democratic values, fundamental rights, and the OECD AI Principles. GPAI prioritizes practical, applied research and applied tools over prescriptive legislative codes, bringing together government representatives, industry leaders, academic researchers, and civil society advocates to co-design guidelines, toolkits, and ethical standards.

The partnership focuses on risk-aware, human-centric oversight that adapts to continuous technological evolution. Rather than enforcing horizontal statutory obligations directly, GPAI develops advisory frameworks—such as its initiatives on agentic AI, responsible AI, and data governance—that member states can voluntarily incorporate into their domestic legislative and policy roadmaps. This approach ensures global interoperability while respecting national sovereignty and regulatory diversity.

Key AI Legislation

Because GPAI is an international partnership rather than a sovereign legislative assembly, it does not enact binding statutory laws or statutory acts. Instead, its regulatory landscape consists of non-binding policy frameworks, summit declarations, and operational initiatives:

• Global Partnership on Artificial Intelligence (GPAI) Framework (2020): The foundational policy structure establishing GPAI's core mandate, working groups, and multi-stakeholder operational model to promote trustworthy and responsible AI grounded in OECD principles.
• New Delhi Declaration (2023): A major ministerial summit declaration reaffirming international consensus on responsible AI, equitable distribution of benefits, and expanded inclusion of the Global South.
• GPAI Agentic AI Systems Framework (2025): A practical policy initiative launched in 2025 by the Tokyo Expert Support Center to establish risk management, transparency, human control, and governance guidelines for autonomous agentic AI systems.

Governance & Enforcement Bodies

GPAI operates through a multi-tiered governance structure designed to balance political leadership with technical research expertise. The highest decision-making authority is the GPAI Council, composed of government representatives from all member states, which determines strategic direction, admits new members, and approves annual work plans. Executive oversight is handled by the Steering Committee, which coordinates day-to-day operations and aligns activities with GPAI's founding principles.

Technical execution is carried out across four thematic Working Groups: Responsible AI, Data Governance, the Future of Work, and Innovation and Commercialization. These working groups are supported by specialized Expert Support Centres, including the Centre of Expertise in Montreal (CEIMIA), the National Institute for Research in Digital Science and Technology (INRIA) in Paris, and the Tokyo Expert Support Center. Administrative and diplomatic coordination is managed by the GPAI Secretariat hosted at the OECD in Paris.

Penalties & Enforcement

As a voluntary international initiative, GPAI possesses no legal authority to impose administrative fines, statutory penalties, civil sanctions, or criminal liabilities on member states or private sector entities. GPAI's governance functions strictly through soft law, peer review, and consensus-building. Its policy frameworks, recommendations, and toolkits carry advisory status and do not create binding legal obligations under domestic or international law.

Compliance and adherence are maintained through membership commitments and reputational incentives. Member states commit to supporting the OECD AI Principles and fundamental human rights. If a member state consistently acts in contradiction to these core values, the GPAI Council maintains the authority to review its status, which can lead to formal suspension or removal from the partnership. Formal regulatory enforcement, judicial appeals, and legal liabilities remain entirely within the jurisdiction of individual member states' domestic legal systems.

Data Protection Framework

GPAI does not promulgate binding data protection laws or statutory privacy regulations. Data privacy obligations within member states remain governed by their respective national legislation. However, GPAI actively addresses data governance through its dedicated Data Governance Working Group, which researches mechanisms to facilitate secure cross-border data sharing while safeguarding individual privacy rights.

Central to GPAI's data governance framework is the concept of 'Data Trusts' and the deployment of privacy-enhancing technologies. These frameworks aim to establish equitable legal and technical structures that protect individual privacy and data security while enabling the collective use of data for public research and social good.

Sector-Specific Rules

GPAI does not issue binding sector-specific regulations for industries such as healthcare, finance, or transportation. Sectoral regulatory mandates remain under the authority of national regulatory agencies within member countries. However, GPAI conducts applied, domain-specific research projects to address targeted challenges across key sectors.

Notable sectoral initiatives include GPAI's 'AI and Pandemic Response' projects, which developed frameworks for leveraging AI in disease tracking while respecting privacy, and research into 'AI and Climate Change'. Additionally, the Innovation and Commercialization Working Group focuses on reducing entry barriers for small and medium-sized enterprises (SMEs) by promoting open-source tools and shared compute infrastructure.

International Alignment

International alignment is the core objective of GPAI. Founded to establish global consensus on responsible AI, GPAI works closely with major international organizations and multilateral forums. Its underlying foundational principles are directly derived from the OECD AI Principles, and its formal integration with the OECD in July 2024 further consolidated global policy alignment.

GPAI maintains active touchpoints with the G7 and G20, directly contributing to initiatives such as the Hiroshima AI Process. It collaborates with UNESCO to align responsible AI work with the UNESCO Recommendation on the Ethics of Artificial Intelligence. Furthermore, GPAI coordinates with regional bodies such as the European Union to ensure its non-binding frameworks remain compatible with binding regional legislation like the EU AI Act.

Future Developments

GPAI's future trajectory is centered on completing and implementing its Agentic AI Systems Framework. Launched by the Tokyo Expert Support Center in February 2025, the draft framework is scheduled for completion in June 2025, followed by expert community review in September 2025, submission to the GPAI Ministerial Council in March 2026, and anticipated adoption by June 2026.

Additional ongoing and planned developments include expanding research into Generative AI governance, safety testing and evaluation frameworks, and strengthening inclusion of emerging economies and the Global South within global AI governance structures.

Enforcement Bodies

Agency	Mandate	Key Powers	Website
GPAI Council	Highest decision-making body of GPAI, setting strategic direction and approving work plans	Admitting new members, approving annual work plans, reviewing member status
GPAI Steering Committee	Executive body overseeing day-to-day operations	Project review and approval, operational oversight
GPAI Secretariat (OECD)	Coordinates partnership efforts, manages international relations, and facilitates communication	Administrative coordination, international relations management
GPAI Expert Support Centres	Provide administrative, financial, and technical support for project execution	Project execution, technical research, framework drafting