GPAI Merges With OECD to Channel AI Research Into Policy Standards

In July 2024, the Global Partnership on AI (GPAI) formally merged with the OECD‘s AI policy work, folding four years of multi-stakeholder research into the OECD’s standard-setting machinery. The integrated body retains the GPAI brand and is designed to give GPAI’s expert outputs a direct path into international policy recommendations.

How GPAI Works

GPAI launched in June 2020 with 15 founding members and had expanded to 29 countries plus the European Union by 2024. The partnership is voluntary and non-binding. Its central mechanism is four multi-stakeholder working groups: Responsible AI, Data Governance, the Future of Work, and Innovation and Commercialisation. Each group draws on experts from government, industry, civil society and academia to produce reports, tools and policy recommendations that member governments can adapt to their national contexts.

The Responsible AI working group has addressed deepfakes and intellectual property in generative AI, and in December 2020 released reports on responsible AI deployment and AI in pandemic response. These outputs carry no enforcement weight but serve as reference points for national regulators, a dynamic familiar to anyone tracking how the EU AI Act draws on international guidance while adding its own binding layer.

AI and Labour Conditions

The Future of Work working group has produced some of GPAI’s most concrete outputs. Its “AI for Fair Work” project, running since 2021, re-evaluates the OECD AI Principles through the lens of workplace applications. In June 2024, the group published the “Fairwork Amazon Report 2024” examining how AI has changed conditions in the warehouse sector. In December 2023, it released “Fairwork AI Ratings 2023, The Workers Behind AI at Sama,” focusing on data annotation workers.

Current projects look at generative AI’s effect on labour markets in South America and the conditions of workers across AI data pipelines. These are assessments, not binding standards, but they carry weight in policy debates where hard data on AI’s labour-market effects remains scarce.

Data Governance and Copyright

The Data Governance Working Group has run technical demonstrations on Privacy-Enhancing Technologies (PETs) in health sector scenarios, testing how sensitive health data can be made available for AI research without compromising individual privacy. Its headline report, “Copyright and Data Protection Rights in Co-Generated Input and Output of Generative AI,” maps the legal landscape across the EU, the United States and Japan. The report does not resolve the underlying tensions but lays out where the frameworks diverge, which has practical relevance for developers building generative AI products across multiple jurisdictions, territory also covered by the broader debate around AI copyright and training data.

Membership and Diplomatic Reach

GPAI’s membership nearly doubled between 2020 and 2024, adding countries including Argentina, Brazil, India and Singapore. Annual summits bring together AI specialists and policymakers to exchange views on recent developments. The partnership’s value proposition rests on that breadth: by convening a diverse set of governments under a shared principles-based framework, it gives smaller economies a seat in AI governance discussions typically dominated by the US, EU and China. Whether the OECD merger strengthens or dilutes that function will depend on how much operational independence GPAI retains and whether its working group outputs continue to reach beyond OECD member states. The UN’s own AI scientific report is navigating a similar tension between breadth and influence.