The Universality Myth: Epistemic Exclusion, Structural Power, and AI Governance in the Global South
This paper argues that the claimed universality of major AI governance frameworks is an ideological myth that masks a structural mechanism of epistemic exclusion, wherein populations in the Global South lack the necessary "contestation literacy" to challenge AI systems, thereby reinforcing the power of deployers through a combination of conceptual resource deprivation, distorted sociotechnical imaginaries, and bureaucratic structural barriers.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine a world where the rules for how artificial intelligence is built and controlled are written in a language most people cannot read. Global leaders and major organizations often claim that these rules apply to everyone, everywhere, designed to protect humanity and the planet. They speak of universal principles that should guide every nation. But a closer look reveals a different reality: for many communities, especially in the Global South—a term used here to describe regions historically left out of the creation of these technologies—these rules are not just difficult to follow; they are invisible. The people living under these systems often lack the basic tools to even recognize that an artificial intelligence system is making decisions about their lives, let alone the ability to question those decisions. This is not simply a matter of needing more education or better internet access. It is a structural gap where the very concepts needed to understand and challenge these systems are missing from the public conversation.
This gap is what researchers Amulya N and Ashwini Prasad S from RV University set out to investigate. They argue that the idea of "universal" AI governance is an illusion that actually helps those in power. The authors suggest that while tech companies and governments are busy teaching people how to use AI products—how to scroll, click, and interact—they are not teaching people how to govern them. There is a crucial difference between knowing how to use a tool and knowing how to hold the maker of that tool accountable. The paper posits that this lack of "contestability"—the ability to see a system as something that can be questioned and regulated—is not an accident. It is a functional part of how power works. When people cannot see the system, they cannot challenge it, and the companies building it face no pressure to change.
To understand how this happens, the researchers looked at two very different places: India and the nations of sub-Saharan Africa. They chose these locations to show that this problem is not unique to one country or one type of government. In India, a nation with its own powerful tech industry and a large government, the problem comes from within. The government passed a major data protection law, but the debate in parliament lasted only about two hours, and the final law included broad exemptions for the state itself. The public, largely unaware of the details of artificial intelligence, did not have the vocabulary or the political pressure to demand better safeguards. The result was a law that looked protective on paper but offered little real protection to ordinary citizens. In sub-Saharan Africa, the problem comes from the outside. Countries there are adopting AI systems and governance rules built in Europe or the United States, often without the local institutions or expertise needed to enforce them. They are told these rules are universal, but because they were designed for different societies, they often fail to protect the local population, leaving them subject to systems they cannot audit or understand.
The researchers found that this invisibility is maintained by two main forces. First, there is a lack of awareness. Many people simply do not know that AI is making decisions about their loans, their welfare benefits, or their safety. Second, there is a form of misinformation, where people do know about AI but only through specific stories that hide the real dangers. Some are told AI is a magical solution that will solve all development problems, while others are told to fear a sci-fi robot takeover. Both stories distract from the actual, quiet harm happening right now: algorithms that discriminate, data that is stolen, and infrastructure that is built without local consent. The paper argues that these stories are not just misunderstandings; they are structural features that keep people from asking the right questions.
One of the most striking findings of the study is that this blindness extends even to the government officials who approve the construction of the data centers that power these systems. In India, a single data center might need up to thirty different permits from various government agencies. An official might approve a building permit or an electricity connection without ever knowing that the building is a data center for an artificial intelligence system. The paperwork does not ask for that information, and the official's job description does not include understanding the broader impact of AI. They are not corrupt; they are simply unaware. They are enabling a massive governance system while remaining completely blind to its existence. This creates a situation where the people who are supposed to be in charge are actually just as excluded from the knowledge as the citizens they serve.
The study challenges the common belief that the solution is simply to teach more people about technology. The authors argue that teaching people how to use AI does not fix the problem if the people in charge have no legal power to stop bad systems, or if the laws themselves are written to exempt the powerful. Adding more voices to the conversation does not help if the conversation is happening in a language those voices cannot speak, or if the rules of the conversation are set by people who do not need to listen. The paper concludes that the claim of universal AI governance is a myth that serves the interests of those who build and deploy these systems. It allows them to claim they are following global standards while operating in a space where no one can check if those standards are actually being met. Until the people living under these systems are given the conceptual tools to see the mirror for what it is, the power to shape the future of artificial intelligence will remain out of their reach.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.