The Global Partnership on AI (GPAI): From OECD Principles to Practice

The Global Partnership on Artificial Intelligence (GPAI), an initiative launched by G7 nations, serves as the action-oriented partner to the OECD Principles on AI. While the OECD established the world’s first intergovernmental standard (the ‘what’ for responsible AI), GPAI’s core mission is to move these high-level principles from theory into practical reality by developing toolkits, supporting collaborative projects, and addressing real-world policy challenges (the ‘how’).

1. The Five Guiding Principles

In our first post we broke down the OECD AI Principles, which GPAI members are committed to upholding, and are structured around five complementary values for the responsible stewardship of trustworthy AI.

1. Inclusive Growth, Sustainable Development, and Well-being: AI should be designed to benefit people, augmenting human capabilities, and contributing to economic and social progress, ensuring these benefits are broadly shared and reduce inequality.
2. Human Rights and Democratic Values (Fairness and Privacy): AI systems must respect the rule of law, human rights, and democratic values throughout their lifecycle. This includes implementing safeguards to ensure fairness, privacy, non-discrimination, and human autonomy.
3. Transparency and Explainability: There must be responsible disclosure regarding AI systems. Stakeholders should be aware when they are interacting with AI, and outcomes should be understandable, allowing affected parties to challenge the results if necessary.
4. Robustness, Security, and Safety: AI systems must be reliable and secure throughout their lifecycle. They should function appropriately under normal and foreseeable adverse conditions, and potential risks must be continuously assessed and managed.
5. Accountability: AI actors (those developing, deploying, or operating AI) must be accountable for the proper functioning of their systems and for respecting the aforementioned principles, based on their specific roles and context.

2. Adoption and Commitment

The OECD AI Principles were adopted in May 2019 by the OECD member countries when they approved the OECD Council Recommendation on Artificial Intelligence. These principles were swiftly acknowledged by the G7 and served as the foundation for the G20 AI Principles, demonstrating rapid international alignment.

• Who Adopted Them: Initially adopted by OECD member nations, the commitment has since expanded to include partners worldwide. Currently, 47 governments have adhered to these principles, including all 44 GPAI member jurisdictions.
• GPAI’s Role: When GPAI was launched in June 2020, its mandate was explicitly defined by a shared commitment to implementing the OECD AI Principles, effectively making them the guiding ethical framework for the entire partnership.

3. Real-World Usefulness and Practical Examples

The true impact of these principles is demonstrated by the practical projects carried out by GPAI’s Expert Working Groups (EWGs), which focus on creating tools and case studies to implement the principles:

• Human Rights & Fairness: The Future of Work WG, in partnership with Fairwork, assessed the working conditions of data annotation companies (like Sama) supplying AI training data. This research led to one company implementing significant changes, such as guaranteeing a living wage and eliminating unpaid unpaid overtime, directly translating the principle of fairness into tangible labor protections.
• Data Governance & Privacy: The Data Governance WG performed technical demonstrations on using PETs in health sector scenarios. This work provides tested frameworks to increase the availability of sensitive health data for AI research (Inclusive Growth/Well-being) while simultaneously respecting individual privacy (Human Rights/Privacy), solving a critical dilemma in medical AI development.
• Transparency & Accountability: The Responsible AI working group develops concrete frameworks, toolkits, and guidelines for the responsible use of Generative AI. This ensures that new, opaque technologies are deployed with clear traceability and responsibility by providing policymakers and developers with practical steps to address emergent issues like deepfakes and intellectual property.
• Robustness & Safety : This initiative gathers diverse use cases of AI in the workplace across different countries. By observing real-world implementation failures and successes, the group identifies systemic risks and develops policy recommendations to ensure AI systems are robust, secure, and safe for workers, for example, by ensuring proper human oversight in automated decision-making processes.
• Fostering a Policy Environment : The Innovation & Commercialization WG focuses on promoting the growth of trustworthy AI. Their work has involved assessing the unique needs of developing countries and providing frameworks for regulatory sandboxes, which allow innovators to test new AI systems under light-touch regulation. This accelerates the deployment of innovative, yet trustworthy, AI tools globally.

GPAI’s true importance is in its proven ability to deliver tangible outcomes for workers, innovators, and citizens worldwide. By moving beyond high-level policy discussions, GPAI’s projects establish it as a critical global entity for implementing responsible AI. Its working groups have achieved measurable results, from successfully compelling a leading AI data provider to guarantee a living wage for its workers, to creating technical frameworks (like Privacy-Enhancing Technologies) that safely unlock sensitive health data for research. GPAI successfully transforms the theoretical foundations of the OECD into coordinated, real-world action, demonstrating that ethical AI stewardship is both achievable and essential in a rapidly evolving global landscape.