Bikram, from Odisha, is trying to preserve his mother tongue, using artificial intelligence (AI), which is significant for a tribe long battling threats to its cultural identity. However, the use of AI in preserving Indigenous knowledge remains debatable, raising concerns about whether AI can protect cultural heritage fairly without undermining the rights of Indigenous communities.
Based on this article and a study, AI is commonly understood as a data-driven sociotechnical system that relies on the large-scale extraction and processing of data, including text, audio, images, and cultural material. Recent research by Xinyi Fu et al., Oladokun et al., and Perera et al. suggest that computational technologies can preserve cultural elements beyond conventional archival practices. For example, virtual reality combined with AI and multimedia gaming has been used to simulate Aboriginal Australian traditions, allowing users to experience cultural stories in ways that static text cannot achieve. Likewise, an interactive gaming platform based on Cañari Indigenous culture in Ecuador demonstrates how digital environments may assist in imparting community knowledge across generations. However, cultural preservation through AI cannot be separated from the knowledge systems on which AI is built. A systematic literature review on Indigenous people and AI highlights a fundamental distinction between Indigenous Knowledge Systems (IKS) and Western Scientific Knowledge Systems (WSK). IKS is relational, intergenerational, and deeply grounded in land, culture, and community, whereas WSK traditionally operates through claims of objectivity, neutrality, and universality. Although such claims present Western knowledge as detached and scientific, scholars argue that WSK is equally shaped by social power and institutional authority. AI reflects Western scientific logic as the primary basis for organizing and validating knowledge. As a result, Indigenous knowledge is often reduced to transferable data detached from the social, cultural, and relational contexts through which it is continuously practiced, interpreted, and sustained within communities.
The dependence of AI systems on open data further intensifies this tension. While openness and reproducibility are often central to technological advancement, the same review notes that open data practices have historically enabled the misuse and exploitation of Indigenous data without consent. In response, the principle of Indigenous data sovereignty has emerged as a critical framework, asserting that Indigenous communities must retain authority over how their knowledge is collected, interpreted, stored, and deployed. This right is also reinforced by Article 31 of the United Nations Declaration on the Rights of Indigenous People, which recognizes Indigenous people’s entitlement to maintain, control, protect, and develop their cultural heritage, traditional knowledge, and intellectual property.
Fairness in AI has gained increasing attention in recent years; numerous studies have shown instances in real-world applications where AI systems have yielded unfair outcomes, particularly impacting marginalized communities. This concern becomes more visible when examining the issue of algorithmic bias. The same review study emphasizes that AI systems frequently reproduce unfair outcomes because biases can enter at every stage of the AI lifecycle, from data collection and model training to evaluation and implementation. In healthcare, for example, machine learning models trained on New Zealand health records produced less accurate predictions for Māori patients than for New Zealand Europeans, while AI screening systems in Australia generated disproportionately high false positives for Aboriginal patients due to unrepresentative training datasets. These examples demonstrate that AI bias is not simply a technical malfunction but a reflection of whose bodies, experiences, and realities are treated as normative during system design. It can be argued that all technical systems are also cultural systems, inevitably encoding the assumptions of those who build them.
This broader pattern of exclusion is echoed in recent literature on AI, creativity, and data governance in the Global South. Payal Arora (2024) notes that although Global South populations contribute significantly to the world’s digital users, labour, and cultural data, they remain structurally underrepresented in the datasets used to train mainstream AI systems, creating what has been termed “data poverty.” Decolonial scholars, therefore, argue that the question is not simply one of representation but one of power: who defines what data is valuable, who decides how creativity is measured, and who benefits from the commercialization of digitally captured cultural expression. In this sense, Indigenous and Global South communities are often positioned as data sources within AI economies rather than as co-authors of technological futures.
The inadequacy of existing legal protections further complicates this issue. Spano and Zhang (2025) argue that conventional intellectual property rights remain fundamentally misaligned with Indigenous knowledge because such frameworks are built around individual ownership, novelty, and fixed authorship, whereas Indigenous knowledge is collective and often orally transmitted across generations. As a result, Indigenous communities frequently find that their cultural knowledge itself remains legally unprotected. This asymmetry enables what can be understood as data colonialism: the extraction, commodification, and monetization of Indigenous cultural resources through digital infrastructures that provide little transparency and even less community control. Concerns surrounding AI models trained on Indigenous language material without clear consent illustrate how technological innovation can continue older colonial patterns under the language of efficiency and accessibility.
To sum it up, AI occupies a deeply paradoxical position in relation to Indigenous knowledge. On one hand, it offers genuine possibilities for preservation and cultural revitalization ; on the other, it remains dependent on data architectures historically shaped by exclusion, appropriation, and unequal ownership. This suggests that the challenge is not whether AI should be used in Indigenous contexts, but under what political and ethical conditions it can be made accountable. Simply increasing Indigenous representation in datasets or digitizing cultural archives is insufficient if Indigenous people remain excluded from decisions regarding collection, access, model training, ownership, and commercial use. Meaningful technological inclusion therefore requires a shift from symbolic participation to structural authority. In this sense, the future of AI in relation to Indigenous knowledge depends less on technological sophistication alone and more on whether governance frameworks are willing to redistribute control, recognize collective cultural rights, and treat Indigenous communities not as passive repositories of data but as sovereign knowledge holders shaping the terms of digital innovation.
Akshata Nete
