From Access to Agency: Why the Global South Needs More Than AI Access

The white paper “From Access to Agency” frames the challenge of AI inclusion as more than getting people online or giving them access to existing tools. Its central argument is that communities in the Global South need the skills, resources and influence to shape AI for their own languages and needs. The case begins with an investment imbalance: AI attracted $258.7 billion in venture capital in 2025—61% of global venture investment—while less than 1% of AI investment went to social impact, according to reporting on the paper’s launch. These figures describe different categories of investment, not a direct comparison of two shares of the same funding pool. 12

Why access is not the same as agency

A technology may be available yet work poorly for people whose languages and contexts are underrepresented in its development. Reporting on the paper cites estimates that more than 90% of the data used to train early large language models came from English-language sources. It also gives a speech-recognition example: error rates below 6% in English and above 60% in another language. The account does not identify the language in that comparison, so the figures should be read as an illustration of disparity rather than a universal measure of system performance. 2

That gap helps explain the paper’s distinction between access and agency. Access means being able to use a tool; agency means having a meaningful role in shaping the systems, skills and governance that determine how it works. The paper’s argument is that local participation matters if AI is to reflect the languages and priorities of the people expected to use it. 24

Where philanthropy fits

The proposal, as described in launch coverage, is for philanthropy to fund areas where markets and governments fall short. That includes helping build local AI capabilities and supporting a shift from being users of AI to having a greater say in its development and governance. The available reporting does not provide enough detail to set out a complete funding plan or specific recommendations for every stage of that shift. 12

This is a targeted role, not a claim that philanthropy can replace public policy or commercial investment. The case is that philanthropic funding can support capacity and social-impact work that may not attract sufficient market funding on its own. 12

What accompanied the UN General Assembly launch

Emirates Foundation and Badr Jafar launched the paper with Oliver Wyman at a UN General Assembly gathering attended by more than 60 leaders from government, business, philanthropy and technology. The convening focused on how philanthropy could help the Global South shape AI rather than simply use it. 216

A separate, related initiative announced around the same UN General Assembly period brought together 60 signatories behind a five-year goal: enabling an estimated 3.4 billion people who speak languages underrepresented in current AI models to use AI in their own language and voice. That goal highlights the scale of the language-access challenge; it is not evidence that the target has already been achieved. 32

Emirates Foundation also announced a partnership with the Responsible AI Future Foundation. The reported focus is helping countries build AI capabilities suited to their own languages, cultures and needs; the available account does not specify the partnership’s terms or outcomes. 34

The practical test: who gets to shape AI?

The white paper’s funding figures and language examples point to a broader question than whether AI tools are available: who has the resources and authority to make them useful and accountable in local contexts? Its answer is to move from access toward agency by supporting local capacity and giving communities a stronger role in AI’s design and governance. 12