EVS presentation · B.Sc. (Hons) Mathematics

Environmental Ethics in the Age of AI

Intelligence now has a physical footprint. Who carries it, and who decides?

Vedant AggarwalRamjas College, University of Delhi

A test from 1949

“A thing is right when it tends to preserve the integrity, stability, and beauty of the biotic community. It is wrong when it tends otherwise.”

Aldo Leopold, A Sand County Almanac

Now run artificial intelligence through the same test.

Whose interests count? What does AI take from the Earth? What does it give back? Who decides?

01 Anthropocentrism

Only humans have intrinsic value. A forest matters because it gives us timber, oxygen and clean water.

02 Biocentrism

Every living thing has a good of its own: a tiger, a neem tree, a microbe.

Albert Schweitzer · Paul Taylor, Respect for Nature (1986)

03 Ecocentrism

Wholes matter too: species, rivers, ecosystems, the land itself.

Aldo Leopold, the land ethic · Arne Næss, deep ecology (1973)

? And AI?

AI has no interests of its own, so it sits outside the circle. Yet it now shapes everything inside it: the energy we burn, the water we draw, and what we choose to protect.

There is no cloud.

“It’s just someone else’s computer.” An old joke among programmers, and literally true.

Where your prompt goes

Every prompt runs in a building like this.

Rows of chips turning electricity into heat, cooled with air and water, around the clock.

485TWh in 2025

950TWh by 2030

7lakh litres

4.2–6.6billion m³ in 2027

62million tonnes

1.2–5.0million tonnes

Promises vs. emissions

Google pledged net zero by 2030, Microsoft carbon negative. Google says its footprint is growing alongside AI; Microsoft points to data-centre expansion.

Google Environmental Report (2026) · Microsoft Environmental Sustainability Report (2026)

The other side of the ledger

The same tool can protect.

Used with care, AI can find patterns in weather, rivers, forests and oceans that no human team could track by hand.

• 40%Cooling

  Less energy for cooling Google’s data centres once DeepMind’s machine learning ran the controls.

  DeepMind, 2016
• 7 daysFloods

  Of flood warning from Google’s Flood Hub. It began as a pilot in Patna, Bihar, and now covers 150 countries.

  Google Research, 2026
• 90%Weather

  Of 1,380 test targets where GraphCast beat the best conventional 10-day forecast.

  Lam et al., Science, 2023
• 1 weekPlastic

  For FAST-PETase, an enzyme engineered with machine learning, to break down PET plastic.

  Lu et al., Nature, 2022

Weighing it up.

Ethics is rarely a yes or a no. It is a set of questions we owe every system before we build it.

01 Who pays? Justice

The benefits of AI reach users everywhere. The heat, water stress and e-waste stay local, often in poorer places with the least say.

02 What do we owe the future? Future generations

“Act so that the effects of your action are compatible with the permanence of genuine human life.”

Hans Jonas, The Imperative of Responsibility (1979)

03 Can we see it? Transparency

Most companies do not publish the energy or water cost of their models. The EU AI Act (2024) now requires providers of general-purpose AI to document energy use.

04 Do we need it? Sufficiency

“Green AI” asks whether the gain is worth the compute. Sometimes a smaller model, or no model at all, is the ethical choice.

Schwartz, Dodge, Smith & Etzioni, Green AI (2020)

Indian roots

An old ethic for a new machine.

Indian traditions treated nature as kin long before anyone said “sustainability”.

1. 1730Khejarli, Rajasthan

   363 Bishnois, led by Amrita Devi and her daughters, gave their lives to stop the felling of khejri trees.

2. 1973Chipko, Uttarakhand

   Villagers hugged trees to stop commercial logging in the Himalaya.

3. 2017Uttarakhand High Court

   Declared the Ganga and Yamuna living legal persons. The Supreme Court later stayed the order.

Constitution of India · Article 51A(g)

“…to protect and improve the natural environment including forests, lakes, rivers and wild life, and to have compassion for living creatures.”

For every Indian citizen this is a fundamental duty. It belongs in the design brief of every AI system too.

A way forward

There is no “away”.

The recycling symbol is a Möbius strip, a surface with only one side. Gary Anderson, a student, designed it in 1970, the year of the first Earth Day. On a one-sided surface there is nowhere to throw things. On one planet, neither.

1. 01Measure

   Publish the energy, water and carbon cost of every model.

2. 02Shrink

   Use the smallest model that does the job. Reuse before you retrain.

3. 03Clean power

   Train when and where the grid is green. Keep data centres out of water-stressed basins.

4. 04Close the loop

   Repair, reuse and recycle hardware. India’s E-Waste Rules (2022) make producers responsible.

5. 05Share fairly

   Give the communities that carry the costs a share of the benefits and a voice in the decisions.

Leopold’s test, updated

“A thing is right when it tends to preserve the integrity, stability, and beauty of the biotic community. It is wrong when it tends otherwise.”

Thank you.

Questions?

Vedant AggarwalB.Sc. (Hons) MathematicsRamjas College, University of Delhi

Sources

1. Leopold, A. (1949). A Sand County Almanac. Oxford University Press.
2. Taylor, P. (1986). Respect for Nature. Princeton University Press.
3. Næss, A. (1973). The shallow and the deep, long-range ecology movement. Inquiry 16.
4. Jonas, H. (1979/1984). The Imperative of Responsibility. University of Chicago Press.
5. International Energy Agency (2025). Energy and AI; (2026) Key Questions on Energy and AI.
6. OpenAI (2025). Usage and energy figures, as reported by Axios and in “The Gentle Singularity”.
7. Google (2025). Measuring the environmental impact of AI inference (Gemini).
8. Ministry of Power (2026). Per-capita electricity consumption, 2024–25. Ember (2026). Yearly electricity data.
9. Li, P., Yang, J., Islam, M. A., Ren, S. (2025). Making AI Less “Thirsty”. Communications of the ACM 68(7).
10. Baldé, C. P. et al. (2024). The Global E-waste Monitor 2024. ITU & UNITAR.
11. Wang, P. et al. (2024). E-waste challenges of generative AI. Nature Computational Science.
12. Google (2026). Environmental Report. Microsoft (2026). Environmental Sustainability Report.
13. Jevons, W. S. (1865). The Coal Question.
14. Schwartz, R. et al. (2020). Green AI. Communications of the ACM 63(12).
15. Evans, R., Gao, J. (2016). DeepMind AI reduces Google data centre cooling bill by 40%.
16. Lam, R. et al. (2023). Learning skillful medium-range global weather forecasting. Science 382.
17. Nevo, S. et al. (2022). Flood forecasting with machine learning models in an operational framework. HESS. Google Research (2026), Flood Hub coverage.
18. Lu, H. et al. (2022). Machine learning-aided engineering of hydrolases for PET depolymerization. Nature 604.
19. Regulation (EU) 2024/1689, the Artificial Intelligence Act.
20. Constitution of India, Articles 48A and 51A(g). E-Waste (Management) Rules, 2022.
21. Mohd. Salim v. State of Uttarakhand (Uttarakhand High Court, 2017).
22. You, J. (2025). How much energy does ChatGPT use? Epoch AI. Luccioni, A. S., Jernite, Y., Strubell, E. (2024). Power Hungry Processing. ACM FAccT.
23. Uptime Institute (2026). Global Data Center Survey. Coastlines: Natural Earth via world-atlas. CO₂ cube: ideal gas at 15 °C, 1 atm.