India must focus on AI and its environmental impact

UPSC Syllabus Topic: GS Paper 3– Science and Technology – Developments and their applications and effects in everyday life.

Introduction

India must focus on AI and its environmental impact as AI adoption expands rapidly across sectors such as health care, agriculture, and industry. While AI-driven growth promises efficiency and innovation, its hidden environmental costs receive limited attention. The development and deployment of AI systems increase energy demand, carbon emissions, water use, and pressure on natural resources. Without recognising and addressing these costs, large-scale AI expansion risks undermining climate goals, water security, and long-term environmental sustainability.

What is the current status of AI use in India?

1. High Adoption Rate: India shows a strong AI adoption with 70% of its firms already running AI projects, compared to the US at 53%.
2. Data Utilization for AI: 91% of Indian companies plan to use their data for training AI models, which is higher than the global average of 62%.

Environmental consequences and sustainability challenges of AI

1. Rising carbon emissions from AI systems: AI development increases carbon emissions due to heavy computing needs during training and deployment. Studies show that training a single large language model can emit nearly 300,000 kg of carbon, adding pressure to climate mitigation efforts.
2. Energy-intensive data centres: Most AI systems operate through large data centres that consume vast amounts of electricity. A single ChatGPT query uses 10 times more energy than a Google search, and in tech hubs like Ireland, data centres may account for 35% of total energy use by 2026.
3. Dependence on fossil fuels: In many regions, data centres still rely on fossil-fuel-based electricity. This links AI expansion directly to greenhouse gas emissions, worsening global warming and undermining climate goals.
4. Power Fluctuations: Unlike conventional computing, AI training involves sudden spikes and drops in electricity use across different phases. These volatile power loads are difficult for grids to absorb and often require diesel-based backup generators, worsening emissions and air pollution.
5. Model Obsolescence: Generative AI models have a short life-cycle, with frequent releases of newer versions. This makes energy spent on training older models largely redundant, while newer models generally require even greater computational and energy inputs.
6. High water consumption: AI infrastructure uses water for construction and cooling. According to United Nations Environment Programme, AI servers could consume 4.2–6.6 billion cubic metres of water by 2027, increasing water scarcity risks in already stressed regions.
7. Electronic waste generation: Data centres generate large volumes of e-waste containing hazardous substances such as mercury and lead. Improper disposal of this waste threatens soil, water, and human health.
8. Resource-intensive hardware production: Producing AI hardware requires huge raw material inputs. Manufacturing a 2 kg computer needs about 800 kg of raw materials, while microchips depend on rare earth elements mined through environmentally destructive practices.
9. Ecological Pressure: Data centres are physical infrastructures embedded in local environments. Their combined electricity use, water consumption, and resource extraction exert indirect but lasting pressures on biodiversity and surrounding ecosystems.

Data gaps and under-reporting of AI impacts

1. Lack of reliable emissions data: Accurate data on AI’s environmental footprint is limited. Estimates vary widely, with the ICT sector contributing 1.8%–3.9% of global GHG emissions, making informed policy difficult.
2. Misleading efficiency claims: Some corporate disclosures understate AI’s environmental costs. For example, a claim that one AI text prompt uses only 0.24 watt-hours of electricity has been criticised for ignoring cumulative and lifecycle impacts.
3. Absence of lifecycle assessment: Most assessments focus only on energy use during operation. Impacts from mining, hardware manufacturing, water use, and disposal remain poorly measured and reported.

Global policy responses and ethical frameworks

1. Ethical recognition of environmental harm: In 2021, UNESCO adopted recommendations recognising AI’s negative environmental and social impacts. Around 190 countries endorsed these non-binding principles.
2. Legislative steps in advanced economies: The European Union and the United States have proposed laws to regulate AI’s environmental effects, including rules on emissions from high-compute activities.
3. Limited global enforcement: Despite ethical guidelines, binding environmental guardrails for AI remain rare. Governments often prioritise innovation and competitiveness over sustainability concerns.

India’s current approach and missing focus

1. Focus on AI as a climate solution: Current discussions in India emphasise how AI can help fight climate change. However, they overlook the environmental costs of building and running large AI models.
2. No dedicated AI impact assessment: India mandates Environmental Impact Assessments under the EIA Notification, 2006, but AI systems are not included. This creates a regulatory gap despite their growing environmental footprint.
3. Need for policy alignment: AI policies are largely disconnected from environmental regulation. This weakens India’s ability to manage long-term ecological risks from digital infrastructure.

What should be done?

1. UNEP recommendations

The United Nations Environment Programme has proposed five measures to limit the environmental impact of artificial intelligence.

1. Impact measurement: UNEP recommends standardised frameworks to assess AI’s environmental footprint across its life cycle. This covers emissions, energy use, water consumption, and material use.
2. Mandatory disclosure: UNEP suggests requiring companies to disclose the direct environmental impacts of AI-based products and services. This improves transparency and policy credibility.
3. System efficiency: UNEP advises improving the energy efficiency of AI algorithms. It also supports water recycling and reuse of hardware components where feasible.
4. Green data centres: UNEP recommends powering data centres with renewable energy and using carbon offset mechanisms. This reduces emissions from high-compute AI activities.
5. Policy integration: UNEP stresses that AI policies should be integrated with existing environmental regulations. This aligns AI development with climate governance.
6. India-specific measures
7. EIA inclusion: The article proposes extending India’s Environmental Impact Assessment framework to AI development. This enables formal scrutiny of environmental risks from AI training and deployment.
8. National standards: India should develop common assessment standards for AI impacts. This requires collaboration between tech firms, think tanks, and civil society groups.
9. Regulatory metrics: India needs consistent sustainability indicators for regulatory use. These metrics should support monitoring, compliance, and informed decision-making.
10. ESG (Environmental, Social, and Governance) reporting: AI-related environmental impacts should be included under ESG disclosure norms. This strengthens corporate accountability in high-compute operations.

Conclusion

AI growth brings clear environmental risks through emissions, energy use, water stress, and resource depletion. India must measure these costs, integrate AI into environmental regulation, and enforce disclosure standards. Without such steps, AI expansion may deepen climate, water, and sustainability challenges instead of supporting long-term development and global environmental goals.

Question for practice:

Discuss how the rapid expansion of artificial intelligence in India is creating environmental challenges and what measures are needed to address its sustainability impacts.

Source: The Hindu