Can AI ever be truly universal? The colonial roots of machine learning

By Bhabani Shankar Nayak*  

Artificial intelligence is no longer a futuristic fantasy. It is rapidly becoming the invisible architect of our daily lives—influencing what we eat, who we befriend, how we heal, and where we work. In the years ahead, AI will not just facilitate human choices; it will increasingly determine them. This inescapable presence presents a stark paradox: AI holds the potential to deepen democracy, decentralise power, and decolonise knowledge, yet it also threatens to entrench the very inequalities it could help dismantle.

The danger lies not in the technology itself, but in the data that fuels it. Most AI systems and their Large Language Models (LLMs) are trained on datasets overwhelmingly drawn from Europe. Consider this: roughly 90 percent of the archival records in the World Digital Library originate from Europe, with barely 10 percent coming from the rest of the world. When AI learns from such lopsided sources, Eurocentric bias is not a bug—it is a feature. These systems do not merely reflect Western worldviews; they elevate them to the status of universal scientific truth, while marginalising indigenous knowledge from Africa, Asia, and Latin America as mere folklore or ethnography.

This epistemic violence is not accidental. The very architecture of AI—its reliance on binary ‘input’ and ‘output’ logic—echoes Cartesian dualism, a framework that historically dismissed non-European knowledge traditions as inferior. By embedding this duality into its core, AI inadvertently reproduces colonial hierarchies, stifling its own emancipatory promise. The goals of democratisation, decolonisation, and even decarbonisation are undermined when the technology that could serve all of humanity instead serves only a narrow, profit-driven elite.

The consequences are already visible. Economically, AI fuels digital capitalism’s platform markets, reducing social relationships to transactional data. Politically, it centralises power in the hands of a few platform capitalists, enabling a quiet authoritarianism that controls everyday life. Culturally, it promotes mass consumerism and standardisation—flattening food, entertainment, and even dreams into homogeneous products. Capitalism promised choice; AI delivers conformity.

Worse still, AI reduces human beings to data points. Our desires, needs, and aspirations are stripped of context, treated as variables in a causal equation that ignores the complex, intergenerational conditions that shape real happiness and peace. While AI has revolutionised productivity and even warfare, it has proven woefully incapable of understanding the richness of human existence.

The solution is not to abandon AI, but to decolonise it. That means radically diversifying the datasets, epistemologies, and governance structures that underpin these systems. AI must be collectively owned, operated, and guided by a genuine plurality of knowledge traditions from around the world. Only then can it fulfil its democratic potential—not as a tool of domination, but as a catalyst for equality, justice, and genuine human flourishing.

---

*Academic based in UK