People have been concerned about environmental issues like deforestation, soil erosion, rising water levels, and air and sound pollution for years. The pollution can be said to have increased after the beginning of the Industrial Revolution as factories for the production of goods on a large scale were set up and emitted harmful gases, and workers lived in squalid conditions. This was 100 years ago, but our environment’s state has not improved. A new technology that has added to the existence of such a condition is the use of AI. It is thought-provoking how something we use via our electronic devices can cause environmental degradation, but it’s true.
First, we need to understand what AI is. Whitni Simpson states that AI is “machines that can learn, reason, and act for themselves” (134). These are basic qualities every human possesses, and this is being transferred to a machine, to an extent. This is the simplest explanation. AI is used in our daily life through models like Chat GPT, which gives answers to prompts put by people, Voice assistants like Siri and Alexa, chatbots which help navigate through different websites, through deep learning apps like amazon prime and Netflix understand what type of content we like and recommend it in our feed. These are just some of the examples of what AI can do.
The core of the harm comes from data centres. Before an AI model is released, data scientists conduct rigorous tests to find its weak points and make corrections in the code. This takes years, and for this to be done, large amounts of electricity need to be used as well as water for cooling. Data centres house the AI, and it is through these centres that people can use its services. In a paper by Patrick K. Lin, they have stated that “The largest data centres require more than 100 megawatts of power capacity, which is enough to power roughly 80,000 U.S. households” to cool these centres large quantities of water is required and they have stated that 3-5 million gallons of water per day is required. The water used for cooling may get polluted and hot, due to which it cannot be reused. This requirement is said to increase over the years as people’s dependence on AI increases, which in turn will lead to an increase in data centres. It can be seen in the increase in data centres from 500,00 in 2012 to 8 million.
Lin has stated that, more than training, Inference uses more electricity and water. It is the process of “Inference is the process of taking an AI model and deploying it onto a device, which will then process incoming data to look for and identify whatever it has been trained to recognize” (17). Basically, this means that it is put into the real world to see if it can do the tasks for which it has been trained. Due to its “repetitious and continuous nature”, it consumes more.
Also, due to the need for constant updating of equipment and rare earth minerals, old equipment is disposed of, mostly not properly, which can lead to the buildup of E-waste in landfills, and the minerals in it can seep through the soil into groundwater.
After the realization of such an effect of AI “In September 2019, workers from 12 tech companies, including Amazon, Facebook, Google, Microsoft, and Twitter, united to form the Tech Workers Coalition and joined the global climate strike, demanding “zero carbon emissions by 2030, zero contracts with fossil fuel companies, zero funding of climate denial lobbying or other efforts, and zero harm to climate refugees and frontline communities.” ” (23, Lin).
Even though AI has created environmental issues, it is also working to mitigate them. It is being used to locate dredging and calculate methane emissions and map climate change on earth.
There are certain measures governments can take/ have taken regarding AI. Proper calculation of the carbon footprint AI leaves needs to be undertaken, and policies for its sustainable use should be formulated. Currently, AI development is following the path of “accuracy over efficiency”, which is not working in terms of sustainability which needs to be rethought, for electricity use renewable energy sources like solar should be used and maybe another type of liquid other than water which gives greater cooling in lesser amount which is reusable.
In Conclusion, the knowledge about the harmful effects which AI is causing should be spread to people, and adequate measures should be taken so that it can be used efficiently and sustainably for the times to come.
References
https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117
https://earth.org/the-green-dilemma-can-ai-fulfil-its-potential-without-harming-the-environment/
https://www.scientificamerican.com/article/ais-climate-impact-goes-beyond-its-emissions/
Lin, Patrick K. “The Cost of Training a Machine: Lighting the Way for a Climate-Aware Policy Framework That Addresses Artificial Intelligence’s Carbon Footprint Problem.” Fordham Environmental Law Review, vol. 34, no. 2, 2023, pp. 1–29
Simpson, Whitni. “A Need to Regulate the Environmental Impacts of Artificial Intelligence (AI): Preserving Clean Water for Humans, Not Robots.” Tulane Environmental Law Journal, vol. 38, no. 1, 2025, pp. 133–48
Naeeni, Sepehr Khajeh, and Nilofar Nouhi. “The Environmental Impacts of AI and Digital Technologies.” AI And Tech in Behavioral and Social Sciences, vol. 1, no. 4, Jan. 2023, pp. 11–18
Article by: Antonio Anthey
.%20These%20are%20basic%20qualities%20every%20human%20possesses,%20and%20this%20is%20being%20transferred%20to%20a%20machine,%20to%20an%20extent.%20This%20is%20the%20simplest%20explanation.%20AI%20is%20used%20in%20our%20daily%20life%20through%20models%20like%20Chat%20GPT,%20which%20gives%20answers%20to%20prompts%20put%20by%20people,%20Voice%20assistants%20like%20Siri%20and%20Alexa,%20chatbots%20which%20help%20navigate%20through%20different%20websites,%20through%20deep%20learning%20apps%20like%20amazon%20prime%20and%20Netflix%20understand%20what%20type%20of%20content%20we%20like%20and%20recommend%20it%20in%20our%20feed.%20These%20are%20just%20some%20of%20the%20examples%20of%20what%20AI%20can%20do.The%20core%20of%20the%20harm%20comes%20from%20data%20centres.%20Before%20an%20AI%20model%20is%20released,%20data%20scientists%20conduct%20rigorous%20tests%20to%20find%20its%20weak%20points%20and%20make%20corrections%20in%20the%20code.%20This%20takes%20years,%20and%20for%20this%20to%20be%20done,%20large%20amounts%20of%20electricity%20need%20to%20be%20used%20as%20well%20as%20water%20for%20cooling.%20Data%20centres%20house%20the%20AI,%20and%20it%20is%20through%20these%20centres%20that%20people%20can%20use%20its%20services.%20In%20a%20paper%20by%20Patrick%20K.%20Lin,%20they%20have%20stated%20that%20“The%20largest%20data%20centres%20require%20more%20than%20100%20megawatts%20of%20power%20capacity,%20which%20is%20enough%20to%20power%20roughly%2080,000%20U.S.%20households”%20to%20cool%20these%20centres%20large%20quantities%20of%20water%20is%20required%20and%20they%20have%20stated%20that%203-5%20million%20gallons%20of%20water%20per%20day%20is%20required.%20The%20water%20used%20for%20cooling%20may%20get%20polluted%20and%20hot,%20due%20to%20which%20it%20cannot%20be%20reused.%20This%20requirement%20is%20said%20to%20increase%20over%20the%20years%20as%20people’s%20dependence%20on%20AI%20increases,%20which%20in%20turn%20will%20lead%20to%20an%20increase%20in%20data%20centres.%20It%20can%20be%20seen%20in%20the%20increase%20in%20data%20centres%20from%20500,00%20in%202012%20to%208%20million.img%20class=size-full%20wp-image-9133%20aligncenter%20src=https://www.bodhini.in/wp-content/uploads/2025/06/Picture1.avif%20alt=%20width=380%20height=331%20/Lin%20has%20stated%20that,%20more%20than%20training,%20Inference%20uses%20more%20electricity%20and%20water.%20It%20is%20the%20process%20of%20“Inference%20is%20the%20process%20of%20taking%20an%20AI%20model%20and%20deploying%20it%20onto%20a%20device,%20which%20will%20then%20process%20incoming%20data%20to%20look%20for%20and%20identify%20whatever%20it%20has%20been%20trained%20to%20recognize”%20(17).%20Basically,%20this%20means%20that%20it%20is%20put%20into%20the%20real%20world%20to%20see%20if%20it%20can%20do%20the%20tasks%20for%20which%20it%20has%20been%20trained.%20Due%20to%20its%20“repetitious%20and%20continuous%20nature”,%20it%20consumes%20more.Also,%20due%20to%20the%20need%20for%20constant%20updating%20of%20equipment%20and%20rare%20earth%20minerals,%20old%20equipment%20is%20disposed%20of,%20mostly%20not%20properly,%20which%20can%20lead%20to%20the%20buildup%20of%20E-waste%20in%20landfills,%20and%20the%20minerals%20in%20it%20can%20seep%20through%20the%20soil%20into%20groundwater.After%20the%20realization%20of%20such%20an%20effect%20of%20AI%20“In%20September%202019,%20workers%20from%2012%20tech%20companies,%20including%20Amazon,%20Facebook,%20Google,%20Microsoft,%20and%20Twitter,%20united%20to%20form%20the%20Tech%20Workers%20Coalition%20and%20joined%20the%20global%20climate%20strike,%20demanding%20“zero%20carbon%20emissions%20by%202030,%20zero%20contracts%20with%20fossil%20fuel%20companies,%20zero%20funding%20of%20climate%20denial%20lobbying%20or%20other%20efforts,%20and%20zero%20harm%20to%20climate%20refugees%20and%20frontline%20communities.”%20”%20(23,%20Lin).Even%20though%20AI%20has%20created%20environmental%20issues,%20it%20is%20also%20working%20to%20mitigate%20them.%20It%20is%20being%20used%20to%20locate%20dredging%20and%20calculate%20methane%20emissions%20and%20map%20climate%20change%20on%20earth.There%20are%20certain%20measures%20governments%20can%20take/%20have%20taken%20regarding%20AI.%20Proper%20calculation%20of%20the%20carbon%20footprint%20AI%20leaves%20needs%20to%20be%20undertaken,%20and%20policies%20for%20its%20sustainable%20use%20should%20be%20formulated.%20Currently,%20AI%20development%20is%20following%20the%20path%20of%20“accuracy%20over%20efficiency”,%20which%20is%20not%20working%20in%20terms%20of%20sustainability%20which%20needs%20to%20be%20rethought,%20for%20electricity%20use%20renewable%20energy%20sources%20like%20solar%20should%20be%20used%20and%20maybe%20another%20type%20of%20liquid%20other%20than%20water%20which%20gives%20greater%20cooling%20in%20lesser%20amount%20which%20is%20reusable.In%20Conclusion,%20the%20knowledge%20about%20the%20harmful%20effects%20which%20AI%20is%20causing%20should%20be%20spread%20to%20people,%20and%20adequate%20measures%20should%20be%20taken%20so%20that%20it%20can%20be%20used%20efficiently%20and%20sustainably%20for%20the%20times%20to%20come.Referencesa%20href=https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117/aa%20href=https://www.unep.org/news-and-stories/story/ai-has-environmental-problem-heres-what-world-can-do-abouthttps://www.unep.org/news-and-stories/story/ai-has-environmental-problem-heres-what-world-can-do-about/aa%20href=https://earth.org/the-green-dilemma-can-ai-fulfil-its-potential-without-harming-the-environment/https://earth.org/the-green-dilemma-can-ai-fulfil-its-potential-without-harming-the-environment//aa%20href=https://www.scientificamerican.com/article/ais-climate-impact-goes-beyond-its-emissions/https://www.scientificamerican.com/article/ais-climate-impact-goes-beyond-its-emissions//aLin,%20Patrick%20K.%20“The%20Cost%20of%20Training%20a%20Machine:%20Lighting%20the%20Way%20for%20a%20Climate-Aware%20Policy%20Framework%20That%20Addresses%20Artificial%20Intelligence’s%20Carbon%20Footprint%20Problem.” emFordham%20Environmental%20Law%20Review/em,%20vol.%2034,%20no.%202,%202023,%20pp.%201–29Simpson,%20Whitni.%20“A%20Need%20to%20Regulate%20the%20Environmental%20Impacts%20of%20Artificial%20Intelligence%20(AI):%20Preserving%20Clean%20Water%20for%20Humans,%20Not%20Robots.” emTulane%20Environmental%20Law%20Journal/em,%20vol.%2038,%20no.%201,%202025,%20pp.%20133–48Naeeni,%20Sepehr%20Khajeh,%20and%20Nilofar%20Nouhi.%20“The%20Environmental%20Impacts%20of%20AI%20and%20Digital%20Technologies.” emAI%20And%20Tech%20in%20Behavioral%20and%20Social%20Sciences/em,%20vol.%201,%20no.%204,%20Jan.%202023,%20pp.%2011–18emstrongArticle%20by:%20Antonio%20Anthey/strong/em){kind=link}