Paper short abstract

This paper studies AI and Sustainability-related policies to examine the socio-ecological implications of AI infrastructures in India and Mexico. It includes interviews with AI policy stakeholders in India and Mexico to understand their perspectives, rationalities and justifications in the process.

Paper long abstract

Growing concerns surround the increasing energy demands and environmental impacts of resource extraction for AI chips, running of AI data centres, and the training of models (Crawford, 2021; Hodgkingson et al., 2024; Lehuedé, 2025). Scholars have brought focus to difficulties with calculating the environmental costs of AI technologies and greenwashing, and there have been efforts to increase awareness about these impacts (Hao, 2024 and 2025; Heikkilä, 2022). Current scholarship at the intersections of AI, sustainability and policy studies has tended to focus on leveraging AI for achieving sustainability goals (Nishant et al., 2020). However, little is known about the experiences and policies that shape the lives of those most affected by new data infrastructures. This is critical because Global South countries bear a larger share of the brunt of the environmental costs associated with data centre and ICT infrastructure development, while reaping fewer benefits from digitisation (UNCTAD, 2024).

Our paper draws on an ongoing research project that examines the socio-ecological implications of AI infrastructure development in India and Mexico, with a focus on AI policies and lived experiences of communities. In this presentation, we specifically focus on AI and Sustainability-related policies, including international, Indian and Mexican policy documents related to AI technologies, as well as adjacent and associated policies (e.g. land acquisition, water resourcing, environmental clearances and labour laws). This is supplemented by semi-structured interviews (n=10) with AI policy stakeholders each in India and Mexico, to understand their perspectives on the sustainability implications of AI technologies and infrastructures.

Paper short abstract

AI governance is often framed through ethical and techno-solutionist models presented as neutral and universal. This paper argues that such approaches reproduce epistemic injustice and digital colonialism, and advances epistemic sovereignty as a decolonial alternative.

Paper long abstract

Artificial Intelligence (AI) governance is dominated by regulatory frameworks that foreground ethical principles. While these approaches are presented as neutral and progressive, this paper argues that they reproduce deep epistemic and structural inequalities when transposed uncritically into Global South contexts. Anchored in ethical determinism and technological solutionism, prevailing AI governance regimes tend to obscure the historical and cultural conditions that shape the production of AI. This paper contends that such approaches perpetuate epistemic injustice and entrench forms of digital colonialism. Wherein Global South societies become sites of data extraction and experimentation while they remain marginal to decision-making and lack equal control over technological design and governance. The asymmetrical distribution of regulatory costs and benefits raises critical questions about whose interests AI governance ultimately serves and whether existing frameworks meaningfully address issues of economic dependency and infrastructural inequality. Against this background, the paper advances the concept of epistemic sovereignty as a normative and political horizon for decolonising AI governance in the Global South. This paper further examines how AI governance is shaped by transnational technology corporations, international standards bodies, and Global North research institutions, while the Global South states, communities, and knowledge producers are marginalised. This mirrors colonial extractive relations where value is generated from Global South resources without meaningful participation or benefit-sharing. In doing so, the paper exposes how tech-solutionist approaches to AI governance obscure power asymmetries and legitimise dependency under the guise of innovation and development.

Paper short abstract

This paper examines AI-driven revenue mobilisation in Kenya, through the integration of telco data with tax systems. It critiques tech-solutionist AI governance under Kenya’s Data Protection Act and AI Strategy, arguing that procedural compliance obscures political, and distributive challenges.

Paper long abstract

This paper critically examines tech-solutionism in AI governance through the empirical case of AI-driven revenue mobilisation in Kenya. Recently, the Kenyan government decided to integrate telecommunications and mobile money transactions data with Kenya Revenue Authority (KRA) systems as a strategy to enhance tax compliance, widen the tax base, and curb revenue leakages (Republic of Kenya, 2023). These initiatives are framed as neutral, efficiency-enhancing technological fixes to structural revenue challenges.

Considering Kenya’s Data Protection Act (2019) and the National Artificial Intelligence Strategy 2025–2030, the paper argues that such solutionist framings obscure the socio-political and distributive dimensions of taxation. While the Data Protection Act provides safeguards against solely automated decision-making and mandates data protection impact assessments (Republic of Kenya, 2019), in practice these mechanisms are operationalised as procedural compliance tools within revenue analytics systems. Application of predictive models trained on integrated data runs the danger of disproportionately targeting informal sector and intensifying surveillance without addressing underlying causes of informality.

The National AI Strategy positions AI on public sector efficiency and economic growth (Government of Kenya, 2024), reinforces a developmental narrative in which technological deployment precedes institutional readiness, and democratic oversight. This dynamic is conceptualised as epistemic displacement, whereby locally grounded understandings of informality, state–citizen trust, and fiscal justice are abandoned in favour of emerging technologies.

By foregrounding revenue mobilisation as a critical site of AI governance, the paper challenges ethical determinism and argues for a context-sensitive approach that treats AI governance as a political and distributive process rather than purely technical.

Paper short abstract

This paper examines how inequalities in AI knowledge production shape governance debates. Using original global co-authorship data, it shows how limited epistemic voice in the Global South risks reinforcing inequitable and developmentally misaligned AI regulation.

Paper long abstract

Recent debates on artificial intelligence governance increasingly emphasise ethical frameworks and regulatory principles as central tools for managing AI’s societal impacts. However, such approaches emerge under conditions of profound inequality in the production and circulation of technological knowledge. This paper examines how uneven participation in AI research shapes the epistemic foundations of AI governance and raises questions about the developmental relevance of emerging regulatory agendas, particularly their alignment with local capacities, priorities, and institutional contexts.

The analysis draws on an original dataset of AI-related research co-authorship from 2013 to 2022, disaggregated across key AI fields including Robotics, Natural Language Processing, Computer Vision, Large Language Models, and AI Safety. Collaboration patterns are analyzed based on countries’ belonging to the Global North or Global South and geopolitical groupings (United States, China, Europe, and Others). The findings reveal a persistent concentration of AI research within high-income countries, alongside highly asymmetric cross-regional collaborations. Field-level analysis further shows that AI Safety remains a comparatively less developed area of AI knowledge production, indicating potential entry points for broader participation by the Global South.

Participation in knowledge production matters because it shapes problem definitions, risk perceptions, and the normative assumptions that inform regulatory agendas. The paper argues that AI governance is shaped not only through formal regulatory processes, but through historically embedded and asymmetrical power relations, with knowledge production serving as a key observable dimension. The paper concludes by outlining alternative pathways for AI governance that foreground epistemic inclusion, capacity-building, and the strengthening of regional research ecosystems.