As policymakers in the Global North prepare to introduce binding horizontal AI regulation, increasing attention is also being paid to the promises and pitfalls of AI for developing countries.  Although much ink has been spilled on the global risks posed by AI to civil and political rights such as privacy, non-discrimination, and freedom of expression, less attention has been paid to the role of economic and social rights (ESRs) in providing a conceptual and framework for understanding and regulating the challenges posed by AI for the Global South. 

ESR issues can crop up at all stages of the AI lifecycle. First, AI for development is hampered by a lack of accurate, up to date data sets in the Global South, especially for underserved communities and languages. As a result, an AI tool to – e.g. enable cancer scans trained on data from the Global North – may create biased outcomes with regard to the right to health when deployed in countries with poor access to diagnostic equipment or different racial and socio-economic compositions. Second, AI for development relies heavily on digital infrastructure controlled by a few large private sector actors based in the Global North. These companies currently operate in a “human-rights free zone” governed largely by self-regulatory efforts, with effective state oversight hampered by a pervasive lack of transparency.

Finally, when integrated into the delivery of public services, AI systems can impact a huge range of ESR’s depending on their deployment context, including the right to work, right to be free from hunger, and right to education. Particular attention must be paid to the rights of marginalised groups and individuals. For example, the right to social protection may be adversely affected where algorithmic decision-making is incorporated into welfare systems without adequate attention paid to digital inclusion and guardrails. These risks are exacerbated by human over-reliance on algorithmic outputs, particularly in jurisdictions where decision-making capacity is limited or digital literacy is low, and where institutional safeguards such as redress mechanisms are lacking.

However, ESRs can also play a transformative, enabling role in AI regulation and governance in the Global South. AI regulatory proposals grounded in human rights (e.g. the Council of Europe’s Draft Framework Convention on AI) often do not deal with ESRs in great detail. An ESR framing for AI regulation in the Global South should reject binary tradeoffs between innovation and regulation, emphasising instead sufficient guardrails to manage negative impacts on marginalised communities, while also supporting governance as an enabler for local AI development in the Global South that supports the Sustainable Development Goals and closes the global AI divide (see UN High-Level Advisory Board’s interim report). An ESR approach to AI regulation should also be developed in consultation with marginalised communities likely to be economically or socially impacted by AI systems, with emphasis on ensuring the ESRs are included within mandatory human rights impact assessment frameworks (see below).

ESRs can also provide a framework for practical AI governance efforts, drawing on existing models such as the Council of Europe’s proposed HUDERAF process (which, although rigorous, does not devote a great deal of attention to ESRs per se). Crucially, an ESR lens requires project teams to ask whether AI is actually needed to achieve the intended development outcome and raise standards of living, from the point of view of rightsholders. There is an urgent need for States and their development partners to begin creating rights-based frameworks to guide AI deployment.  Although frameworks such as the World Bank’s Environmental and Social Framework deal with issues closely related to ESRs, these are not necessarily tailored to the unique considerations faced by AI deployment, and are often limited in scope (e.g. the ESF only applies to Investment Project Financing). The World Bank’s Human Rights, Inclusion and Empowerment Umbrella Trust Fund is currently funding a project seeking to deliver recommendations on a toolkit to assess human rights risks arising from Bank-funded operational projects involving AI (disclaimer: the author is a consultant on this project). Ultimately, an ESR-based approach rejects artificial distinctions between sustainable development and human rights, allowing the creation of AI for development that improves standards of living for all and enables full enjoyment of citizens’ fundamental rights.