Equitable Access to AI for Climate Adaptation in Vulnerable Regions

By Moumita Das

This article benefited from excellent research assistance provided by Toby Qin, East Chapel Hill High School.

Introduction

The escalating crisis of climate change presents a profound challenge to global ecosystems and societies, with extreme weather events and gradual environmental shifts intensifying at an alarming rate.¹ These impacts are not felt uniformly across the globe; rather, regions in the Global South are disproportionately affected.² The Global South broadly comprises Africa, Latin America and the Caribbean, Asia (excluding Israel, Japan, and South Korea), and Oceania (excluding Australia and New Zealand).⁴¹ Furthermore, developing countries in the Global South face even greater risks from the effects of climate change, which collectively threaten their socio-economic stability.² Unfortunately, the Global South, despite contributing the least to the greenhouse gas emissions driving climate change, bears the brunt of its devastating consequences.

Artificial intelligence (AI) holds the theoretical potential to revolutionize climate adaptation efforts through its capacity to process and analyze vast datasets with great speed and accuracy, which can offer enhanced understanding and help to predict future climate impacts.⁴ This capability extends to improving the accuracy and speed of weather forecasts and providing crucial early warnings for a range of disasters.³ AI applications in climate adaptation span diverse areas, including the prediction of floods, wildfires, and droughts, as well as the optimization of resource management in critical sectors such as agriculture and water.⁴ The theoretical potential of AI in climate adaptation is significant, but equitable access and deployment is necessary in order to actually apply this potential to vulnerable populations.

Climate change necessitates a dual approach encompassing both mitigation (reducing greenhouse gas emissions) and adaptation (adjusting to the impacts of a changing climate).³ Although much of the initial AI innovation in this domain has focused on mitigation strategies, there is a growing need to accelerate the development and equitable deployment of AI for climate adaptation in the Global South.³ Strengthening the ability of vulnerable regions to adapt to both current and anticipated climate events by using actionable climate insights is paramount in minimizing harm from adverse effects emerging from these events.³ 

Addressing the global challenge of climate change requires robust financial mechanisms and a clear understanding of individual responsibilities. International climate finance plays a critical role in supporting climate change control in developing countries.⁸ While developed countries made a commitment to mobilize 100 billion USD annually by 2020 to aid climate action in developing nations, this target has not been consistently met.⁸ 

The principle of Common but Differentiated Responsibilities (CBDR), a key concept in international environmental law, holds that all countries share an obligation to address climate change, but not equally. Developed nations, having contributed more to historical greenhouse gas emissions and possessing greater economic resources, should take on a larger share of the burden when it comes to managing the results of climate change.⁹ As such, CBDR is based on the expectation that wealthier nations should provide financial and technological support to developing countries, both for climate mitigation and adaptation.⁹ Climate finance and the CBDR principle both underscore the need for developed nations to actively support the Global South in their climate adaptation endeavors, including the equitable access to AI technologies.

2. An Overview of AI Applications in Climate Adaptation

Artificial intelligence is increasingly being applied across various sectors to bolster climate adaptation efforts. Key areas where AI is making significant contributions include early warning systems for disasters, enhancing agricultural resilience, and optimizing water management practices.

In the realm of early warning systems for disasters, AI is proving to be a transformative technology. Its ability to analyze vast amounts of data from diverse sources—such as weather sensors, satellites, and even social media—enables more accurate and timely prediction of extreme weather events.¹⁰ AI algorithms can process live data from sensors and drones to provide real-time situational awareness during disasters, detecting anomalies like rising water levels, increased temperatures, or seismic patterns that may indicate impending floods, wildfires, or earthquakes.¹⁰ The World Economic Forum’s FireAId project, for example, uses AI to map high-risk wildfire areas and potential areas of fire development, almost in real time.³ Google's AI-powered Flood Hub provides flood forecasts and early warnings in numerous countries, including several in Africa, demonstrating the potential for widespread impact.¹⁹ These advancements are crucial for providing vulnerable communities with the precious time needed to evacuate and take protective measures, ultimately saving hundreds of millions of lives.

AI is also playing a vital role in increasing agricultural resilience against climate change. AI in precision agriculture allows farmers to optimize yields through data-driven insights and automation.¹³ It analyzes a multitude of data, including soil composition, weather forecasts, crop health (gleaned from satellite imagery and IoT sensors), and pest patterns, to recommend optimized fertilization strategies, guide irrigation and planting schedules, and detect diseases or nutrient deficiencies early.⁶ For example, AI-driven robots can differentiate crops from weeds for reduced herbicide use, and AI-powered irrigation systems can monitor soil moisture and adjust water distribution in real time to minimize water waste.¹⁴ 

Optimizing water management is another area where AI is being effectively applied for climate adaptation. By analysing of datasets related to weather patterns, water availability, water usage, and population growth projections, AI assists governments in optimizing water distribution networks.⁷ It can predict water demand, detect leaks in water infrastructure, and optimize pumping schedules to minimize water waste and enhance network performance.⁷ In wastewater treatment, AI enables real-time adjustments to aeration rates by monitoring pollutant levels and flow rates, reducing energy use while maintaining effective treatment. Companies like IDRICA and Xylem Vue offer AI-powered platforms for smart water management that use predictive analytics, sensor integration, and real-time monitoring to enhance water distribution, reduce operational inefficiencies, and promote long-term sustainability.¹⁵ These systems not only reduce costs for utilities but also help to build climate-resilient infrastructure by ensuring more reliable access to clean water and reducing exposure to water-related climate risks. These advancements in water management illustrate a broader trend: AI is becoming central to climate adaptation strategies across diverse sectors. IBM’s flood mapping platform, Microsoft’s water-focused agricultural tools in Kenya, and ESCAP’s multi-hazard AI systems further demonstrate how localized applications are scaling globally.¹²

A wide range of additional efforts highlight this growth. ESCAP's Asia Pacific Risk and Resilience Portal employs predictive AI to improve multi-hazard risk modeling in developing countries.⁴ In India, the Farmer.Chat app provides vital agricultural information in local languages.⁴ BrainBox AI is contributing to adaptation by helping building owners reduce energy costs.⁴ Weather forecasting is being revolutionized by companies like Tomorrow. io, using satellites and AI.¹¹ The agricultural sector benefits from AI solutions offered by companies like CropIn, IBM Watson, and Blue River Technology.¹³ A notable collaboration between Oxford University, WFP, and ICPAC is using AI for more accurate weather forecasting in East Africa.¹⁶ Google's Project Green Light aims to reduce emissions by optimizing traffic flow.¹⁷ Collectively, these initiatives exemplify the expanding role of AI in addressing the challenges associated with climate adaptation.

3. Barriers to Equitable Access and Deployment of AI for Climate Adaptation

Despite these promising applications, several barriers hinder equitable access and deployment of these tools in vulnerable and resource-constrained regions. These barriers include data availability, technical capacity, infrastructure limitations, contextual relevance, and ethical considerations.

A critical challenge is the data divide, the gap between those who have access to high-quality data and those who do not. The data divide can negatively skew development across countries and is therefore a serious issue that needs to be addressed.⁴² AI models are data-driven, requiring high-quality, high-frequency datasets such as remote sensing imagery, local climate patterns, and infrastructure maps.³ Yet, in many low-income regions, particularly within the Global South, such data may be missing, inaccessible, or of insufficient quality.³ The ownership of climate-related data is also a complex issue, involving various stakeholders like individuals, research institutions, and private companies, leading to ambiguities in consent and usage rights, especially in cross-border contexts.²⁰ Furthermore, datasets available in the developing world are often low quality and unstandardized, which directly impacts the accuracy and reliability of AI algorithms trained on them.²¹ The scarcity of adequate, accessible, and relevant data in vulnerable regions is a major impediment to harnessing the full potential of AI for climate adaptation in these areas

Technical capacity gaps pose another substantial barrier to equitable access and deployment. The development, customization, and interpretation of AI models demand specialized expertise that is often lacking in the regions that are most vulnerable to climate change.³ Educational institutions and research programs in many developing countries lack the necessary resources to adequately train and retain skilled scientists, data analysts, and computational experts in the fields of AI and climate modeling.¹⁸ This shortage of local technical expertise significantly hampers innovation and limits the effective implementation of AI-driven solutions in these regions.¹²

In addition, smaller local organizations often face significant financial hurdles in implementing and maintaining AI systems, further widening the gap in technical capabilities.¹⁹ The concentration of AI expertise and resources in wealthier nations thus creates a significant capacity deficit, preventing vulnerable regions from fully leveraging AI to address their specific climate adaptation needs.

Infrastructure deficiencies constitute a further impediment. Limited computational resources, inadequate internet connectivity, and insufficient institutional support in many vulnerable regions act as major constraints on local AI innovation and adoption.³ The development and operation of sophisticated AI climate models requires substantial computational infrastructure, including high-performance computing facilities, advanced software, and specialized hardware, which are predominantly located in developed nations.¹⁸ Basic infrastructure challenges, such as unreliable access to electricity and insufficient internet penetration, are prevalent in many parts of the Global South.¹²

Contextual misfit presents another key challenge. AI models are often trained on data from specific regions and socio-environmental contexts, which may not accurately reflect the diverse characters of other areas, particularly in the Global South.³ Climate change impacts and socio-economic conditions vary significantly across different regions, necessitating tailored adaptation strategies rather than universally applicable solutions.²² For instance, agricultural practices common in the Global South, such as polyculture, may not be adequately represented in typical agricultural datasets used to train AI models, leading to potentially irrelevant or ineffective recommendations.²³ This lack of contextual fit can result in AI models that do not generalize well across different socio-environmental settings, thereby limiting their utility in vulnerable regions.

Ethical considerations are paramount in ensuring equitable access and deployment of AI for climate adaptation. There is a significant risk that AI tools, if not developed and deployed responsibly, could exacerbate existing inequalities and promote "top-down" adaptation strategies that do not align with the needs of vulnerable communities.³ Algorithmic bias, stemming from biased training data, can perpetuate and even amplify societal prejudices, leading to unfair or discriminatory outcomes for marginalized groups.² Concerns about data privacy and potential misuse of personal information also arise, particularly in the context of disaster management.²⁶ An over-reliance on AI-driven systems could lead to policy interventions that overlook the needs of certain social groups, especially those with limited access to digital technologies.²⁷ Addressing these ethical considerations is crucial to ensure that AI serves as a tool for promoting climate justice rather than reinforcing existing vulnerabilities.

4. Economic Impacts and Uneven Development Trajectories

The unequal distribution of AI-driven climate resilience technologies has major economic consequences, especially for developing regions. Without equitable access, these regions risk falling further behind, as limited technological capacity can exacerbate existing inequalities and delay economic development in the Global South.

The global climate transition necessitates substantial investments, especially within emerging markets and developing economies, where AI is uniquely positioned to catalyze clean growth and generate new economic opportunities.²⁸ AI's capacity to enhance productivity, create possibilities for employment, and stimulate inclusive growth in the context of climate adaptation is considerable.²⁸ However, the unequal distribution and application of AI on a global scale risk further widening the economic chasm between wealthier and poorer nations.⁵ 

The impacts of AI are likely to both aggravate and alleviate the conditions of uneven development.²⁴ The increasing centralization of AI production can lead to significant power imbalances, positioning the Global South primarily as consumers of AI technologies rather than as equal partners in their creation and governance.²⁹ Consequently, many countries in the Global South face the distinct risk of being left behind in the ongoing digital revolution, while simultaneously becoming integrated into a global AI supply chain that potentially extracts value for the benefit of foreign entities, with limited accountability to local needs and priorities.²⁹ This dynamic can perpetuate a new form of technological dependency, where the Global South remains reliant on AI solutions conceived and developed in the Global North, thereby hindering the development of self-driven, contextually appropriate solutions.

Scaling AI tools, including Large Language Models (LLMs), in scenarios involving low-resource languages presents both significant challenges and potential opportunities for equitable access to climate adaptation technologies. The development and deployment of LLMs and other AI tools in regions where low-resource languages are prevalent are hampered by the aforementioned problem of data scarcity and quality³⁰ The inherent bias and language imbalance in training data can lead to performance disparities and the introduction of cultural biases in AI outputs for these languages.³¹ 

These limitations are compounded by technical challenges that further hinder the scalability of AI in low-resource contexts. Tokenization, for example, can be less efficient for low-resource languages.³² However, there are promising approaches, including the development of regional multilingual models and monolingual models specifically tailored for individual low-resource languages, which can help to address these limitations.³⁰ AI can play a role in language preservation and the creation of digital resources for these languages, potentially bridging the communication gap.³³ 

5. The Uneven Burden of AI Development and Deployment

The development and deployment of artificial intelligence for climate adaptation incurs an uneven burden in relation to the environmental and energy costs associated with the necessary infrastructure.

These costs, associated with AI infrastructure, such as data centers and high-performance computing facilities, are substantial and growing.²⁵ Data centers, which form the backbone of AI operations, are massive consumers of electricity, potentially increasing the reliance on fossil fuels and contributing to greenhouse gas emissions.³⁴ The training of large AI models, essential for many climate adaptation applications, requires immense computational power and can result in a significant carbon footprint.²⁵ Beyond energy consumption, AI infrastructure also demands vast quantities of water for cooling purposes, which can exacerbate water scarcity issues in already vulnerable regions.³⁴  The manufacturing and disposal of AI hardware contribute to the growing problem of electronic waste and necessitate the extraction of critical minerals, often through unsustainable mining practices.³⁴ These concerns underscore the importance of ensuring that the deployment of AI for climate adaptation is itself sustainable and does not inadvertently worsen the climate crisis it aims to address.

The burden of the environmental and energy costs associated with AI development and deployment is often disproportionately placed on certain communities. Low-income communities, for example, may be more likely to reside near power plants or data centers that support AI infrastructure, thus bearing a greater share of the associated environmental pollution and health risks.³⁵ Air pollution emanating from data centers can lead to significant public health issues, including respiratory illnesses and premature deaths, with estimated annual costs running into billions of dollars.³⁵ This raises the critical question of environmental justice, prompting calls for technology companies to adequately compensate the communities most affected by the environmental consequences of their data processing centers.³⁵ The concept of community responsibility and the principle of providing full compensation for damages caused by AI systems become particularly relevant in this context.³⁶ Given that AI’s benefits are global while its environmental harms are focused locally, there is a strong case for frameworks that hold developers accountable for these harms.

6. Pathways Towards Inclusive and Just AI for Climate Adaptation

Achieving inclusive AI for climate adaptation, particularly in vulnerable regions, requires a concerted effort guided by equity, ethics, and sustainability. Several pathways can be pursued to ensure that AI serves as a tool for empowerment and resilience for all.

One such framework is Sheila Jasanoff's concept of "technologies of humility".³⁷ This concept advocates for researchers and policymakers to recognize the inherent limitations and uncertainties in scientific and technological endeavors.³⁷ It emphasizes the need to integrate the technical capabilities of AI with critical ethical considerations, prompting us to ask fundamental questions about the purpose, potential harms, beneficiaries, and our understanding of these technologies.³⁸ In the context of AI for climate adaptation, adopting "technologies of humility" means proceeding with caution, actively seeking diverse perspectives, especially from vulnerable communities, and ensuring that AI solutions are socially relevant, ethically sound, and contribute to climate justice rather than exacerbating existing inequalities.

Initiatives like Oxford’s AI4Climate Initiative and Google AI for Social Good projects are actively working to promote the localized and equitable access to AI tools for climate adaptation noted in the technologies of humility pathway. The Oxford AI4Climate Initiative focuses on leveraging AI for public good, specifically addressing climate challenges through projects like AI-based weather forecasting in East Africa, developed in collaboration with the World Food Programme (WFP) and the IGAD Climate Prediction and Applications Centre (ICPAC).¹⁶ This initiative aims to improve forecast accuracy in resource-constrained regions without requiring additional costly infrastructure.¹⁶ It also explores the potential of AI in areas like energy efficiency, wildfire management, flood response, and integration with indigenous knowledge.³⁹ Google AI for Social Good pursues projects demonstrating AI's positive impact across various societal challenges, including climate change, accessibility, healthcare, and economic opportunity.⁴ Their efforts include AI-powered tools for flood forecasting, wildfire tracking, traffic light optimization to reduce emissions, and climate resilience initiatives in Africa.⁴ Through programs like the AI Impact Challenge and the Generative AI Accelerator, Google.org provides funding and technical support to organizations developing AI solutions for social good, including climate action.⁴⁰ These initiatives represent a growing commitment to harnessing AI for climate adaptation in vulnerable regions, often emphasizing the development of localized solutions and fostering collaborations with local partners. To achieve truly equitable access and lasting impact, sustained and scaled efforts are essential.

7. Conclusions

This analysis reveals a complex landscape surrounding the equitable access to and deployment of AI tools for climate adaptation, particularly in vulnerable and resource-constrained regions. While AI offers tremendous potential to enhance our ability to predict, prepare for, and respond to the impacts of climate change, significant barriers currently impede its equitable application. Data divides, technical capacity gaps, infrastructure deficiencies, contextual mismatches, and ethical concerns all contribute to a situation where the benefits of AI may not reach those who need them most.

The economic implications of this unequal access are profound, with the potential to further entrench uneven development trajectories in the Global South. The challenges of scaling AI tools in low-resource language scenarios further exacerbate these disparities, highlighting the need for targeted efforts to ensure linguistic inclusivity. Moreover, the environmental and energy costs associated with AI development and deployment raise concerns about sustainability and environmental justice, necessitating careful consideration of the burden placed on certain communities.

However, pathways towards a more inclusive and just AI ecosystem for climate adaptation do exist. Embracing the principles associated with "technologies of humility," as articulated by Sheila Jasanoff, can guide the ethical and responsible development and deployment of AI. Initiatives like Oxford’s AI4Climate Initiative and Google AI for Social Good are making strides in promoting localized and equitable access, but these efforts need to be scaled and sustained.

To truly harness the power of AI for climate adaptation in vulnerable regions, a multi-faceted approach is required. This includes strategic investments in local expertise and digital infrastructure, the promotion of open-source AI solutions, the fostering of collaborations between diverse stakeholders, and the establishment of equitable data governance and ethical frameworks. Special attention must be paid to addressing the unique challenges of low-resource languages and ensuring that AI development itself adheres to principles of sustainability and environmental justice. Ultimately, upholding the principle of Common but Differentiated Responsibilities requires developed nations to take a leading role in supporting the Global South through technology transfer and financial assistance, ensuring that AI becomes a powerful tool for building climate resilience for all.

Reference List

1.     AI for Resilient Infrastructure in Climate Vulnerable Regions Future → Scenario, accessed April 26, 2025, https://prism.sustainability-directory.com/scenario/ai-for-resilient-infrastructure-in-climate-vulnerable-regions-future/ 

2.     UNCTADstat Classifications - United Nations Conference on Trade and Development (UNCTAD), accessed April 26, 2025, https://unctadstat.unctad.org/EN/Classifications.html 

3.     How AI can help the world prepare for climate adaptation, accessed April 26, 2025, https://www.weforum.org/stories/2022/11/how-artificial-intelligence-can-prepare-us-for-climate-adaptation/ 

4.     Leveraging AI for climate adaptation: A new era for resilience - PreventionWeb.net, accessed April 26, 2025, https://www.preventionweb.net/news/leveraging-ai-climate-adaptation-new-era-resilience 

5.     What opportunities and risks does AI present for climate action? - Grantham Research Institute on climate change and the environment - LSE, accessed April 26, 2025, https://www.lse.ac.uk/granthaminstitute/explainers/what-opportunities-and-risks-does-ai-present-for-climate-action/ 

6.     www.datasciencecentral.com, accessed April 26, 2025, https://www.datasciencecentral.com/precision-agriculture-powered-by-ai-for-climate-resilient-crops/#:~:text=Weather%20Data%3A%20AI%20integrates%20real,diseases%20or%20nutrient%20deficiencies%20early.

7.     AI in water management: Balancing innovation and consumption | White & Case LLP, accessed April 26, 2025, https://www.whitecase.com/insight-our-thinking/ai-water-management-balancing-innovation-and-consumption 

8.     Using Machine Learning to Track International Climate Finance, accessed April 26, 2025, https://www.climatechange.ai/blog/2022-11-09-climate-finance 

9.     The colonial past and current climate privileges: Who pays for the crisis? - xqthenews, accessed April 26, 2025, https://xqthenews.com/es/el-pasado-colonial-y-los-actuales-privilegios-climaticos-quien-paga-la-factura-de-la-crisis/ 

10.  AI in Disaster Management: How Artificial Intelligence is Saving Lives in Emergencies, accessed April 26, 2025, https://www.hashstudioz.com/blog/ai-in-disaster-management-how-artificial-intelligence-is-saving-lives-in-emergencies/ 

11.  How AI Is Being Used to Respond to Natural Disasters in Cities | TIME, accessed April 26, 2025, https://time.com/7171445/ai-natural-disaster-cities/ 

12.  How AI is transforming climate risk management in developing countries | illuminem, accessed April 26, 2025, https://illuminem.com/illuminemvoices/how-ai-is-transforming-climate-risk-management-in-developing-countries 

13.  Precision agriculture powered by AI for climate-resilient crops - DataScienceCentral.com, accessed April 26, 2025, https://www.datasciencecentral.com/precision-agriculture-powered-by-ai-for-climate-resilient-crops/ 

14.  AI in Agriculture: Smarter Crops, Healthier Livestock, Better Yields - Cloud Security Alliance, accessed April 26, 2025, https://cloudsecurityalliance.org/blog/2025/02/10/ai-in-agriculture-smarter-crops-healthier-livestock-better-yields 

15.  Five key areas in which Artificial Intelligence is set to transform water management in 2025, accessed April 26, 2025, https://www.idrica.com/blog/artificial-intelligence-is-set-to-transform-water-management/ 

16.  New AI-led science initiative will help protect communities hit by climate change in East Africa | University of Oxford, accessed April 26, 2025, https://www.ox.ac.uk/news/2024-06-27-new-ai-led-science-initiative-will-help-protect-communities-hit-climate-change-east 

17.  AI for sustainability - Google AI, accessed April 26, 2025, https://ai.google/applied-ai/sustainability/ 

18.  Equitable Access to AI Climate Modeling Data → Scenario - Prism → Sustainability Directory, accessed April 26, 2025, https://prism.sustainability-directory.com/scenario/equitable-access-to-ai-climate-modeling-data/ 

19.  AI for Climate Action - BMZ Digital.Global, accessed April 26, 2025, https://www.bmz-digital.global/wp-content/uploads/2024/09/AI-for-Climate-Action-Paper.pdf 

20.  Data Ownership in the AI-Powered Integrative Health Care Landscape - PMC, accessed April 26, 2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC11615554/ 

21.  AI Governance: A Unique Challenge for the Global South - Global Diplomacy Lab, accessed April 26, 2025, https://www.global-diplomacy-lab.org/blog/ai-governance-a-unique-challenge-for-the-global-south/ 

22.  Assessing the Potential of AI–ML in Urban Climate Change Adaptation and Sustainable Development - MDPI, accessed April 26, 2025, https://www.mdpi.com/2071-1050/15/23/16461 

23.  Farming on Screen - Resilience.org, accessed April 26, 2025, https://www.resilience.org/stories/2024-03-05/farming-on-screen/ 

24.  The Backroads of AI: The Uneven Geographies of Artificial Intelligence and Development | Request PDF - ResearchGate, accessed April 26, 2025, https://www.researchgate.net/publication/337202066_The_Backroads_of_AI_The_Uneven_Geographies_of_Artificial_Intelligence_and_Development 

25.  The US must balance climate justice challenges in the era of artificial intelligence, accessed April 26, 2025, https://www.brookings.edu/articles/the-us-must-balance-climate-justice-challenges-in-the-era-of-artificial-intelligence/ 

26.  AI in natural disaster management - Ultralytics, accessed April 26, 2025, https://www.ultralytics.com/blog/ai-in-natural-disaster-management 

27.  Artificial Intelligence for Climate Security: Possibilities and Challenges - SIPRI, accessed April 26, 2025, https://www.sipri.org/sites/default/files/2023-12/2312_sipri_policy_report_ai_for_climate_security_1.pdf 

28.  AI's role in the climate transition and how it can drive growth - The World Economic Forum, accessed April 26, 2025, https://www.weforum.org/stories/2025/01/artificial-intelligence-climate-transition-drive-growth/ 

29.  Advancing a More Global Agenda for Trustworthy Artificial Intelligence, accessed April 26, 2025, https://carnegieendowment.org/research/2024/04/advancing-a-more-global-agenda-for-trustworthy-artificial-intelligence?lang=en 

30.  Mind the (Language) Gap: Mapping the Challenges of LLM Development in Low-Resource Language Contexts | Stanford HAI, accessed April 26, 2025, https://hai.stanford.edu/policy/mind-the-language-gap-mapping-the-challenges-of-llm-development-in-low-resource-language-contexts 

31.  Multilingual LLMs: Progress, Challenges, and Future Directions - Prem AI Blog, accessed April 26, 2025, https://blog.premai.io/multilingual-llms-progress-challenges-and-future-directions/ 

32.  8 Challenges Of Building Your Own Large Language Model - Labellerr, accessed April 26, 2025, https://www.labellerr.com/blog/challenges-in-development-of-llms/ 

33.  AI in Language Preservation: Safeguarding Low-Resource and Indigenous Languages, accessed April 26, 2025, https://www.welocalize.com/insights/ai-in-language-preservation-safeguarding-low-resource-and-indigenous-languages/ 

34.  Explained: Generative AI's environmental impact | MIT News, accessed April 26, 2025, https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117 

35.  Air Pollution and the Public Health Costs of AI - Division of Engineering and Applied Science, accessed April 26, 2025, https://www.eas.caltech.edu/news/air-pollution-and-the-public-health-costs-of-ai 

36.  Full article: Rethinking compensation in light of the development of AI, accessed April 26, 2025, https://www.tandfonline.com/doi/full/10.1080/13600869.2024.2324554 

37.  Sheila Jasanoff: We need technologies of humility - Rathenau Instituut, accessed April 26, 2025, https://www.rathenau.nl/en/digitalisering/sheila-jasanoff-we-need-technologies-humility 

38.  Technologies of Humility - P2P Foundation Wiki, accessed April 26, 2025, https://wiki.p2pfoundation.net/Technologies_of_Humility 

39.  AI for Climate Change - Oxford Insights, accessed April 26, 2025, https://oxfordinsights.com/projects/ai-for-climate-change/ 

40.  AI for the Global Goals 2023 - Google, accessed April 26, 2025, https://globalgoals.withgoogle.com/globalgoals/

41.  UNCTADstat Classifications - United Nations Conference on Trade and Development (UNCTAD), accessed May 12, 2025, https://unctadstat.unctad.org/EN/Classifications.html

42.  Toward Bridging the Data Divide - World Bank Blogs, accessed April 26, 2025, https://blogs.worldbank.org/en/opendata/toward-bridging-data-divide