Ethical Guidelines for Use of AI in Environmental Research

Abstract

The use of Artificial Intelligence (AI) is becoming increasingly prevalent across environmental research. AI enables benefits in biodiversity monitoring, early warning systems, environmental modeling, and research optimization. However, AI imposes environmental costs, including high water consumption, high energy needs, air pollution, and impacts on public health. This paper examines the ethical paradox of AI’s potential for environmental good combined with its current environmental harms. It analyzes the relevant ethical tradeoffs through a variety of ethical lenses: Deontology, Justice, Utilitarianism, and Daoism. Finally, it proposes guidelines to help environmental researchers responsibly balance AI’s environmental impact while maximizing its environmental benefits.

Overview

Benefits of Using AI in Environmental Research:

Before the widespread adoption of Artificial Intelligence (AI), many researchers relied on other statistical methods. This transition is reflected in Konya and Nematzadeh’s (2024) review of AI in environmental-related fields, which found that “Among the studied articles, over 65% of environmental tasks that demonstrate interest in using AI tools initially relied on conventional statistical and mathematical models” (p. 1). These findings illustrate the increasing prevalence of AI in environmental research. This paper focuses specifically on generative AI, whose role is expanding in environmental research due to its ability to process large datasets and generate complex outputs. This focus is particularly important given that generative AI has a greater environmental impact than other AI forms.

AI is now widely used to analyze complex climate datasets, track air quality, and conduct lifecycle analyses to measure a product’s footprint (UN Environment Programme, n.d.). These capabilities allow researchers to optimize data analysis and improve accuracy.

This growing use of AI is reflected in a wide range of environmental research applications. For example, AI has been used to improve estimates of Arctic sea ice loss, enhance air quality monitoring, detect key fish species, reduce energy consumption (Konya and Nematzadeh, 2024),and model oceanic biogeochemical processes (Liniger et al., 2025). The increasing role of AI is also evident in current research trends. The Ecological Society of America’s 2025 Baltimore conference featured numerous AI-driven studies, which included work on dengue prediction, ocean conservation using genomics and remote sensing, and machine learning analyses of wetland biodiversity (Ecological Society of America Public Affairs, 2025). These examples demonstrate how AI is becoming central to advancing environmental knowledge.

In addition to its research applications, AI is also driving measurable environmental improvements. For instance, Google reports that its geospatial products, including Google Earth, Solar API, Google Maps’ fuel-efficient routing, and Green Light, have helped lower GHG emissions (tCO2e) by over 18 million metric tons (Google Earth AI, n.d.). In the renewable energy sector, AI is projected to increase load factors of solar photovoltaics and wind by up to 20% (The London School of Economics and Political Science, 2025). These findings suggest that AI can contribute not only to understanding environmental issues but also to addressing them.

In the future, AI may become more sustainable as technologies improve. For example, according to Robb (2025), one innovation under development is a cooling chip that could improve the system efficiency by around 25%. While traditional cooling systems require considerable energy and water to maintain safe temperatures for the chip, newer approaches, such as microfluidic cooling, remove heat more effectively at its source. Since this method reduces the need for intensive cooling, it can lower both energy and waste consumption in data centers. Consequently, continued advancements could allow AI to maximize its benefits in environmental research while minimizing its environmental footprint.

Environmental Costs of AI

Despite these benefits, the rapid expansion of generative AI has introduced significant environmental tradeoffs. Artificial Intelligence training and processing, along with much of the internet traffic, is performed by computer servers that are housed in locations such as data centers. The infrastructure needed to support AI systems requires substantial resources and can harm communities.

Electricity

While much of the environmental research that utilizes AI seeks to better understand the environment, AI can also lead to unintended environmental harm through high energy consumption. Data centers, which power AI systems, require large amounts of electricity. In 2023, data centers accounted for 4.4% of total US electricity consumption (Shehabi et al, 2024).This percentage is projected to reach 12% of U.S. electricity generation by 2028 (SolarTech, 2025). This trajectory underscores the scale of energy required to support these technologies.

However, energy demands vary depending on how the technology is used. For example, Luccioni et al. (2024) found that the most efficient models for processing 1000 questions used enough energy to power a 25W incandescent light bulb for 5 minutes, while using the least efficient models to generate 1,000 HD images could require enough energy to charge a smartphone about 70 times. This also illustrates how different tasks vary in environmental footprint. For instance, generating images is more energy-intensive compared to generating text, and generating videos is even more demanding. This raises additional concerns regarding the environmental costs.

To power the generation of various types of content, data centers rely on constant power (O’Donnell and Crownhart, 2025). However, the environmental impact of this energy use depends heavily on how that electricity is generated. According to O’Donnell and Cownhart (2025), “In 2024, fossil fuels, including natural gas and coal, made up just under 60% of the electricity supply in the US. Nuclear accounted for about 20%, and a mix of renewables accounted for most of the remaining 20%.” Since many data centers are located in regions heavily reliant on fossil fuels (O’Donnell and Crownhart, 2025), AI’s energy demand may contribute to increased greenhouse gas emissions. While AI supports environmental research, such as climate monitoring, it may simultaneously exacerbate the problems it is trying to address.

In addition to high energy consumption, AI infrastructure also places growing demands on another critical resource: water.

Water

One major environmental impact of AI is the large volume of water consumed by data centers, which often conflicts with community water access. Data center servers produce substantial heat from their electricity usage and require significant water use for cooling (Barringer, 2025).

Additionally, water is also consumed indirectly to generate electricity (Ahmad, 2024). It is estimated that large data centers can consume up to 5 million gallons daily, comparable to the water demand of a town of 10,000 to 50,000 inhabitants (Yanez-Baruevo, 2025).

Cases of impacts to local communities illustrate how data centers sometimes correlate with reduced local water availability. In Newton County, Georgia, for example, nearby data center operations may have affected water quality, leaving some residents with discolored or non-potable water to use (Tan and Chambers, 2025). Similarly, in La Esperanza, Mexico, the expansion of data centers has been linked to water shortages, with some residents experiencing reduced access to water (Arandia and Dib, 2025). These examples highlight how data center water use can directly affect both water availability and quality for communities.

This impact is further compounded by the fact that much of this water cannot easily be reused. Water used by data centers is often treated with chemicals to prevent corrosion and bacterial growth, which may make it unsuitable for other purposes such as human consumption or agriculture (Olson et al., 2024). This limits opportunities for direct recycling while increasing the overall water demand.

AI’s large water demand is particularly severe in water-stressed regions, where the majority of new and planned data centers since 2022 are located (Nicoletti et al., 2025). Supporting this concern, one study on data center environmental footprints found that around 20% of server water use is linked to moderately to highly water-stressed watersheds (Siddik et al, 2021). Although water-use efficiency has improved at the data center level, the total data center water demand continues to grow in water-scarce regions due to industry expansion (Silverstein, 2026).

Furthermore, AI’s water consumption can decrease water availability in local ecosystems, which may disrupt species that depend on these resources. For example, some aquatic life is sensitive to water levels.

Although technologies such as dry and hybrid cooling exist and can help mitigate water consumption, not every data center adopts these due to increased costs, reduced efficiency in warm climates, or the need for regulatory approval to recover expenses (Silverstein, 2026). Nevertheless, these technologies offer potential pathways for reducing water consumption in the future.

Beyond resource consumption, AI infrastructure also contributes to environmental pollution (eg,air pollution), which directly impacts public health.

Air Pollution and Public Health

AI infrastructure also creates health impacts. Data centers often use diesel backup generators that release pollutants such as NOx and PM2.5 (Liu et al., 2025, p. 5). In a recent report produced by the policy nonprofit Next 10, U.C. Riverside researchers report that in 2023, California data centers were estimated to emit a total of 2.38 million short tons of carbon, with resulting public health costs estimated at $155.44 million (Liu et al., 2025, p. 5). Other researchers have reported similar findings regarding data centers and their effects on communities: “Communities living in close proximity (i.e., one mile) to EPA-regulated data centers have higher air pollution burdens compared to the national median (i.e., 50th percentile of air pollution)” (Hampton and Nost,2025). Thus, data centers and their emitted air pollution place local (host) communities at a higher health risk.

Application of Ethical Lenses

This paper seeks to use various ethical lenses to help frame sustainable AI guidelines.

Deontology

A Deontological lens focuses on rights and responsibilities. As described by Velasquez et al (2014), a right is “a justified claim on others,” founded on an acknowledged societal standard. Some rights are formally written in law, while others are moral rights that may not be legally enforced, yet are still widely accepted (Velasquez et al, 2014). Correspondingly, people have duties or responsibilities to uphold rights (Velasquez et al, 2014).

The concepts of rights and responsibilities are reflected in our duty to the environment. Humans have a duty not to cause harm. Since humans created many environmental problems, they also have a responsibility to help repair the damage caused. This applies to AI and its environmental impacts. In this context, environmental protection is not only about preserving ecosystems but also about safeguarding fundamental rights, such as the right to clean air and health.

This lens can be used to evaluate the environmental impacts of AI on rights. For example, AI datacenter infrastructure can contribute to increased air pollution. This pollution has direct public health consequences, which infringe on the right to health. A rights-based approach does not allow these harms to be overlooked simply because AI provides benefits. Instead, environmental researchers have the responsibility to balance these rights when using technologies.

A Deontological perspective also extends beyond human-centered concerns. Environmental rights should also be considered. These include the rights of non-human nature, such as ecosystems and species. These rights are typically philosophically founded upon the intrinsic value of life, meaning that nature has worth independent of its utility for humans. Environmental ethicist Paul Taylor (1981) discusses how granting legal protection to nature can reflect respecting inherent worth: “To grant them [referring to non-human entities] legal protection could be interpreted as giving them legal entitlement to be protected, and this would be a means by which a society that subscribed to the ethics of respect for nature could give public recognition of their inherent worth” (Taylor, 1981, 218).

Given these considerations, environmental researchers play an important role in upholding both human and environmental rights. Additionally, their role in society is especially impactful in problem-solving and undoing the damage caused by humans. As their work often aims to help the environment and reduce negative human impact, they have the responsibility to ensure the use of AI does not cause additional damage. This role is also reflected in a duty to use AI in ways with as minimal negative environmental impacts as possible.

Justice

Another perspective, the justice lens, emphasizes fairness and states that if there is no ethically relevant difference, groups should be treated equally (Velasquez et al., 2014). In the context of environmental research, AI has the potential to both support and undermine this principle. For example, AI can help ensure equality and equity in ways such as helping researchers understand the environmental impacts on communities disproportionately affected by pollution and other impacts of a changing climate. This can help improve awareness and create solutions to help communities. However, the usage of AI can also reinforce existing inequalities. For example, Kapor Foundation (2025) research has shown that California data centers are disproportionately located in communities that are already facing environmental threats, such as poor air and groundwater quality, and hazardous waste. This placement worsens existing environmental and public health injustice for communities and ecosystems, rather than alleviating them.

These inequities are especially visible at local levels. To ensure fairness for everyone, including ecosystems, environmental researchers must understand impacts not only on a global level but also on local communities. A justice lens asks who bears the burdens of the externalities and which groups have a voice. This is particularly relevant with AI-driven water consumption. As discussed above, data centers may stress local water supplies. This sometimes leads to reduced access or degraded water quality for local communities. These impacts demonstrate how the cost of AI infrastructure is often unevenly distributed, which raises justice concerns.

In addition to focusing on who is affected, procedural justice focuses on the decision-making process. Schulz et al. (2025) define procedural justice as requiring the active involvement of the public in decision-making. This means that communities affected by environmental research and infrastructure should have a meaningful impact on shaping the research process and its outcomes. For instance, in participatory science, the public is involved in multiple research stages (United States Environmental Protection Agency, 2026). Sibyl Divers, a researcher at Stanford University, further explains that community-based research involves building partnerships with local communities and learning from their perspectives (Peacock, 2025). Community-based research engages communities in the scientific process and also the process of collecting their own data, as in the case of the Local Environmental Observer Network, which is a crowdsourced mapping tool (United States Environmental Protection Agency, 2026).

Researchers can use community-based participatory research or communicate findings through community engagement to reduce the likelihood that some communities will bear most of the burden of environmental research and try to distribute benefits fairly for everyone.

While community-based participatory research is helpful, communities may still bear an unfair burden. This creates an opportunity for environmental researchers to impact communities differently by conducting environmental research on the local environmental impacts of AI or other community issues, which helps to make the community members agents of self-defense. This approach demonstrates respect, acknowledgement, and transparency for community members.

Utilitarianism

According to Ethics Unwrapped (University of Texas at Austin) (n.d.), “Utilitarianism holds that the most ethical choice is the one that will produce the greatest good for the greatest number.”

Utilitarianism focuses on the costs and benefits of an action and determines if a choice is morally right based on its outcomes (Ethics Unwrapped, n.d.). Within this framework, the tradeoff for using AI is complicated and involves harms and benefits spread over various time scales. As a result, this introduces an ethical conflict between weighing immediate impacts and future outcomes.

In the short term, AI’s harms often create negative environmental impacts. As discussed earlier, the high energy and water usage associated with AI are already affecting communities. As these impacts are occurring now, they are easier to identify and measure compared to long-term benefits.

In the long term, AI has the potential to significantly advance environmental research and sustainability efforts. For example, AI supports data analysis and optimization, including improved climate modeling and biodiversity modeling (Konya and Nematzadeh, 2024). As previously mentioned, these benefits are outlined in current research trends shown in the Ecological Society of America’s 2025 Baltimore conference. This provides an immediate benefit of better analysis of vast datasets. However, the outcomes of this research often take time to materialize. There are often various lag times between identifying an environmental concern and taking action on it (Hocherman et al., 2025). Research benefits may take years to emerge while environmental harms are often immediate (Hocherman et al., 2025). Furthermore, it is difficult to track, understand, and even quantify the potential impacts of one research project, as research often builds on previous studies and may not be utilized until later.

However, generative AI may also have a role in creating more immediate sustainability enhancements. For example, generative AI tools can reduce emissions and environmental impacts when used.

Given the broader tension between short and long-term effects, the Utilitarian perspective places weight on whether AI’s long-term benefits can offset its immediate environmental costs. This lens would also weigh the long-term potential for AI to completely mitigate its own harms and be a more sustainable technology. There is a possibility that researchers will use the technology to one day undo all the harm it has caused. At the same time, it is also important to note that some environmental damage from AI could be difficult to reverse, and relying on future solutions does not address immediate environmental costs. In weighing the environmental benefits and harms, a more sustainable present-day AI that can reverse its current externalities might tilt the scales differently.

Finally, key Utilitarian considerations include understanding whether the use of generative AI is necessary or whether there are better alternatives. Researchers can evaluate the urgency of the situation the research is studying, how many people might benefit from the research, and the extent of the negative impact caused. They also need to consider how short-term harms are distributed and who is affected. The utilitarian calculus informs the guidelines listed below that aim to help mitigate harmful effects.

Daoism

Some perspectives, such as the Chinese ethics lens of Daoism, use a more holistic approach that emphasizes humans’ connection to nature. Viewed through a Daoist lens, being closer to nature provides a better perspective on navigating challenges and encourages people to take better care of the natural world (Cline et al., 2024). Environmental researchers often already hold this perspective, since some of them live and work within the natural world through field work and direct observation of environmental change. Seeing environmental impacts firsthand reinforces Daoist principles that humans are not separated from nature, but are also a smaller part of the whole (Cline et al., 2024). From this perspective, the use of tools such as AI should reflect and respect this interconnectedness.

This holistic view also emphasizes how nature should be understood and valued. Cline et al. (2024) state that Daoists believe that nature should not only be viewed as “knowledge to be gained (e.g., a list of names),” but also “as something to be savored, valued, enjoyed and Treasured.”

Chinese ethics moves away from purely mathematical thinking and toward understanding transformation and interrelatedness (D’Ambrosio, 2023). This challenges approaches that treat environmental systems as isolated data points or purely technical problems. It suggests that environmental researchers need an understanding beyond the models and statistical analysis they are creating. Researchers must consider who these technologies and research outcomes affect.

Environmental research varies in scope, and while some studies focus on human impacts and statistical analysis, others extend beyond these approaches to consider broader ecological relationships. When using AI, researchers need to have a deeper intention to care for the natural world, not just for efficiency.

In addition, Daoist philosophy emphasizes continuous change and transformation. D’Ambrosio (2023) notes that ecosystems and human societies are always changing. Due to this, AI should be understood as connected to a larger system rather than an isolated technology. Researchers should consider AI’s broader implications. A Daoist approach calls for the intention to use AI not only for innovation but also to support sustainability across the entire ecological system.

Lenses Summary

The ethical lenses discussed above provide perspectives on AI’s impact on the environment and on how to help use it responsibly. Deontology focuses on how AI affects rights by upholding or undermining them, especially for vulnerable groups, and also acknowledges conflicts between rights. The Justice lens examines how AI disproportionately impacts certain groups and identifies solutions such as community-based research and partnerships with directly affected groups. A Utilitarian calculus focuses on maximizing benefits and minimizing harms by weighing AI’s harms against benefits and adjusting how researchers use the technology. Daoism takes a holistic approach and emphasizes that humans are part of nature and thus have a responsibility to protect it. Instead of a cold Utilitarian calculus, this approach recognizes that AI is part of a larger system and requires researchers to use AI in ways to support nature, including human well-being.

Professional and Expert Input

In developing the guidelines as part of this project, I sought input from professional environmental researchers, engineers, environmental ethics philosophers, and industryrepresentatives working on sustainability-related projects. For their input, I met with each of these professionals for roughly 30 minutes and asked them a series of set questions with varied follow-up questions. Additionally, I wanted to gain a deeper understanding of AI infrastructure from engineers.

In my discussions, the professionals I met with explained the difference between previous versions of AI and generative AI. They noted that other types of machine learning have lower environmental impacts compared to Generative Artificial Intelligence, which researchers have only recently begun to use more widely. Most of the professionals mentioned that they have not used it directly in their research yet, but they do use it for tasks like summarizing articles to assist with literature reviews, polishing writing, brainstorming ideas, and creating assignments for teaching. While most of these tasks are not directly related to research, many researchers are considering using AI in the future for research tasks such as image processing. Similarly, some appreciated AI’s ability to process data faster and mentioned that they liked the ability to train artificial intelligence with their own datasets.

Throughout many of the discussions, the interviewees emphasized the need for increased transparency. They highlighted a lack of available data center operations data, in particular, related to the amount of water used. Without accurate and accessible data, they noted, it is difficult to make informed decisions.

Some professionals also noted that no technology is absolutely neutral. All technology has an environmental or moral impact. In the case of AI, while it is possible to use it for good, clear guidelines are needed. These guidelines will help balance different values, address concerns across stakeholder groups, and better protect the environment. This connects to using education to spread awareness of AI’s environmental impact.

Need for Guidelines

AI’s unintended environmental harms demonstrate the need for guidelines to better inform environmental research. While AI offers powerful tools for understanding and addressing environmental challenges, environmental researchers must also consider its trade-offs. By adopting these practices, researchers can help ensure that AI contributes positively to environmental sustainability, rather than exacerbating existing challenges.

Guidelines

AI helps accomplish a wide range of tasks and may be necessary in some cases. For example, Ag Stephens, who works for the Centre for Environmental Data Analysis and National Centre for Atmospheric Science (n.d.), explains, “‘Using AI can be like using a super-efficient assistant. If it means spending less time and energy on routine work, the environmental impact might actually be lower overall” (National Centre for Atmospheric Science, n.d.). Therefore, AI should not be completely avoided. However, researchers should understand the environmental impacts of their actions and how to best use available tools effectively. Environmental researchers can use AI when necessary and should focus on using the tool while mitigating harms.

The following guidelines outline responsible and sustainable AI usage. These guidelines draw on the ethical lenses discussed above, and build on existing guidance for AI use, including recommendations from National Centre for Atmospheric Science (n.d.), which provides AI usage tips for researchers in “Conscious Use of AI: Practical Tips and Thoughts from environmental researchers”; the Massachusetts Institute of Technology’s (2025) “Guidance for use of Generative AI tools”; and the University of Florida’s (n.d.) “Guidance for Researchers.” The guidelines are further informed by my discussions with professionals in environmental research and sustainability-related fields.

1. 1. Researchers Should Be Conscious of When to Use AI. Researchers should use alternatives if they are equally effective or more sustainable. Rather than defaulting to AI, they should evaluate whether it is necessary for a given task. Planning tasks in advance helps reduce extraneous AI usage. The National Centre for Atmospheric Science (n.d.), for example, recommends asking, “‘Would I do this differently without AI?’” If the case is yes, and AI does not provide significant benefits, researchers should avoid using AI for that task.
   2. Researchers Should Use a Variety of Techniques to Make AI More Sustainable.
      1. Using effective prompting can reduce inefficient usage. Dr. Dan Hodson, a research scientist, notes that, “‘In some cases, using ChatGPT is ~10x faster than writing the initial draft code myself (with many internet searches). But it does depend on how precisely I word the prompt’” (National Centre for Atmospheric Science, n.d.). Researchers should use more precise prompts to decrease the number of queries needed to attain a result and lower the overall computational demand.
      2. Choosing the appropriate model can also reduce environmental impact. Researchers should use smaller or more specialized models to help reduce a task’s carbon footprint. This shift may already be occurring: The CEO of AI model-hosting platform HuggingFace, Samuel Axon, has mentioned that he expects a shift away from LLMs to a variety of small models that solve a task(Axon, 2025). This approach ensures that researchers use only the computational power needed.
      3. Various tools help schedule computation during cleaner electricity times. Some software forecasts when low-carbon electricity is available and the corresponding times. For example, the Climate Aware Task Scheduler is a Python package made by MIT that uses carbon intensity forecasts to schedule a time to run computations on low-carbon electricity (Bartholomew et al., 2025). It also provides information on how much carbon was saved (Bartholomew et al., 2025). These tools help translate more abstract environmental responsibility into quantifiable decisions.
   3. Researchers Should Use Tools to Improve Transparency on Environmental Impacts.¹ Researchers should ensure transparency regarding AI’s usage and its impact on the carbon footprint, and they can use carbon and energy calculators to track those impacts. Examples of such tools include the ML CO2 Impact calculator by Schmidt et al and AI Energy Consumption Calculator by AI Energy Calculator. There is also an AI EnergyScore implemented by HuggingFace (Luccioni et al., n.d.), which suggests energy-efficient AI models. To promote transparency, researchers must document their own AI usage and other tracked environmental emissions. This involves citing where AI was used and can also include reporting results from calculators used in research papers.
   4. Researchers Should Engage with Local Communities that are Impacted by AI. Researchers can support communities through policy engagement and community partnerships, including community education efforts. This responsibility can be difficult in practice, as researchers may not know where data centers supporting their own AI usage are located and who may specifically be impacted. In such cases and in general, engagement can take broader forms, such as supporting transparency and advocating for more sustainability in technology. This may involve engaging the community  more in the research process, such as in community-based participatory research.

AI Usage Note:

Generative AI tools were used as part of this project to assist in identifying sources (eg, journal articles), improve sentence clarity and grammar, and help organize the paper.

 

1 The impact of these tools has yet to be thoroughly studied, and may depend on a variety of factors such as the scale of the project.

 

Work Cited

Ahmad, R. (2024, March 4). Engineers often need a lot of water to keep data centers cool. American Society of Civil Engineers; American Society of Civil Engineers.

AI Energy Calculator. (n.d.). AI Energy Consumption Calculator; AI Energy Calculator. Retrieved March 28, 2026.

Athanassoulis, N. (n.d.). Virtue Ethics. Internet Encyclopedia of Philosophy; Internet Encyclopedia of Philosophy. Retrieved March 28, 2026.

Axon, S. (2025, November 19). “We’re in an LLM bubble,” Hugging Face CEO says—but not an AI one. Ars Technica; Condé Nast. 

Barringer, F . (2025, April 8). Thirsty for power and water, AI-crunching data centers sprout across the West. Stanford University; Bill Lane Center for the American West at Stanford University.

Bartholomew, S. L., Colling, L., Dasgupta, A., Greenberg, A. J., Lannelongue, L., Lestang, T., Martinez-Ortiz, C., Sauzé, C., Walker, A. M., & Ward, A. S. (2025). CATS: The Climate Aware Task Scheduler. The Journal of Open Source Software, 10(111), 1–5. The Journal of Open Source Software. 

Brennan, A., & Lo, N. Y. S. (2021, December 3). Environmental Ethics. Stanford Encyclopedia of Philosophy; The Metaphysics Research Lab, Department of Philosophy, Stanford University. 

Cline, E., J, P ., & Tagliabue, C. M. (2024, April 7). Early Ecological Ethics: A Look at Chinese Environmental Thought. Common Home; The Earth Commons. 

D’Ambrosio, P . (2023). AI Ethics Beyond the Anglo-Analytic Approach: Humanistic Contributions from Chinese Philosophy. Asian Studies, 11(3), 17–46. ResearchGate. 

Ethics Unwrapped. (2025). Utilitarianism. Ethics Unwrapped; The University of Texas at Austin.

Google Earth AI. (n.d.). Google Earth AI. Google AI; Google. Retrieved March 28, 2026.

Google Sustainability. (2025, June). 2025 Environmental Report. Google Sustainability; Google.

Green Algorithms. (n.d.). Green Algorithms; Green Algorithms. Retrieved March 28, 2026.

Hagendorff, T. (2022). A Virtue-Based Framework to Support Putting AI Ethics into Practice. Philosophy & Technology, 35, 1–24. SpringerNature Link.

Hampton, L. M., & Nost, E. (2025, October 28). Communities Close to EP A-Regulated Data Centers Face Heightened Air Pollution. Environmental Data and Governance Initiative;Multiplier.

Hocherman, T., Trop, T., & Ghermandi, A. (2025). Time lags in environmental governance: A critical review. Ambio, 54(12), 2042–2059. PubMed Central. 

Jiménez Arandia, P ., & Dib, D. (2025, August 17). The Backyard of AI. EL P AÍS 50; Ediciones EL P AÍS.

Keller, J. B., Donoghoe, M., & Perry, A. M. (2024, January 29). The US must balance climate justice challenges in the era of artificial intelligence. Brookings; Brookings.

Konya, A., & Nematzadeh, P . (2024). Recent applications of AI to environmental disciplines: A review. Science of the Total Environment, 906, 1–14. ScienceDirect.

Li, P ., Yang, J., Islam, M. A., & Ren, S. (2025). Making AI Less “Thirsty”: Uncovering and Addressing the Secret Water Footprint of AI Models. ArXiv (Cornell University), 1–10.

Liniger, G., Sharp, J. D., Takeshita, Y., & Johnson, K. S. (2025). Two Decades of Increase in Southern Ocean Net Community Production Revealed by BGC-Argo Floats. GlobalBiogeochemical Cycles, 39(8), 1–19. Advancing Earth and Space Sciences.

Liu, R., Wu, Z., & Ren, S. (2025, November 12). An Assessment of California Data Centers’ Environmental and Public Health Impacts | Next 10. Next10.org; Next 10.

Luccioni, A. S., Pistilli, G., Sefala, R., & Moorosi, N. (2025, April 1). Bridging the Gap: Integrating Ethics and Environmental Sustainability in AI Research and Practice. ArXiv.org; arXiv. 

Luccioni, S., Gamazaychikov, B., Hooker, S., Pierrard, R., Strubell, E., Jernite, Y., & Wu, C.-J. (2024). Light bulbs have energy ratings — so why can’t AI chatbots? Nature, 632(8026), 736–738. Nature. 

Luccioni, S., Pierrard, R., Jernite, Y., Moutawwakil, I., Mitchell, M., & Gamazaychikov, B. (n.d.). AI Energy Score . Hugging Face. Retrieved March 28, 2026.

Marrinan, C., Hinton, L., Goins, R., & Koshy, S. (2025). The Unequal Burden of Data Centers: An Examination of Environmental and Public Health Impacts on Communities in California. Kapor Foundation; Kapor Foundation.

Massachusetts Institute of Technology. (2025, February 19). Guidance for use of Generative AI tools. MIT Information Systems & Technology; Massachusetts Institute of Technology.

Musizvingoza, R. (2024, June 7). AI as a Catalyst for Sustainable Progress: Ensuring Information Integrity. United Nations University Institute in Macau; United NationsUniversity Institute in Macau.

National Centre for Atmospheric Science. (n.d.). Conscious use of AI: Practical tips and thoughts from environmental researchers. National Centre for Atmospheric Science; National Centre for Atmospheric Science. Retrieved March 28, 2026.

Nicoletti, L., Ma, M., & Bass, D. (2025, May 8). AI Is Draining Water From Areas That Needs It Most. Bloomberg.

O’Donnell, J., & Crownhart, C. (2025, May 20). We did the math on AI’ s energy footprint. Here’s the story you haven’t heard. MIT Technology Review; MIT Technology Review. 

Olson, E., Grau, A., & Tipton, T. (2024, July 19). Data centers draining resources in water-stressed communities. The University of Tulsa; The University of Tulsa. 

Peacock, C. (2025, July 16). What does it mean to do “community-based research”? Stanford Report; Stanford University. 

Public Affairs. (2025, July 17). Media Tip Sheet: Artificial Intelligence at #ESA2025. Ecological Society of America; Ecological Society of America. 

Robb, D. (2025, November 17). Microfluidics Enhances AI Chip Performance. IEEE Spectrum; IEEE. 

Santa Clara University. (2014, August 1). Calculating Consequences:The Utilitarian Approach to Ethics. Santa Clara Markkula Center for Applied Ethics; Santa Clara University.

Schmidt, V ., Luccioni, A. (Sasha), Lacoste, A., & Dandres, T. (n.d.). Machine Learning CO2 Impact Calculator. ML CO2 Impact. Retrieved March 28, 2026.

Schulz, A. J., Rempel Fisher, G. M., Nadeem, A., Savoie, K., Israel, B. A., Landrum, T., Stanley, E., & Sampson, N. (2025). Procedural Justice in Environmental Decision Making: An Analysis of Public Participation Language in State Level Cumulative Impact Legislation. Environ Justice, 1–25. National Library of Medicine.

Shehabi, A., Newkirk, A., Smith, S. J., Hubbard, A., Lei, N., Siddik, M. A. B., Holecek, B., Koomey, J., Eric, M., & Sartor, D. (2024). 2024 United States Data Center Energy Usage Report. Lawrence Berkeley National Laboratory, 1–79. eScholarship.

Siddik, M. A. B., Shehabi, A., & Marston, L. (2021). The environmental footprint of data centers in the United States. Environmental Research Letters, 16(6), 1–11. ResearchGate.

Silverstein, K. (2026, January 11). America’ s AI Boom Is Running Into An Unplanned Water Problem. Forbes.

SolarTech. (2025, October 2). How Much Electricity Does a Data Center Use? Complete 2025 Analysis. SolarTech; SolarTech.

Tan, E., & Chambers, D. (2025, July 16). Meta Built a Data Center Next Door. The Neighbors’ Water Taps Went Dry. The New York Times.

Taylor, P . W. (1981). The Ethics of Respect for Nature. Environmental Ethics, 3(3), 197–218.

Tedesco, M. (2025, July 2). AI and Climate Justice: Balancing Risk With Opportunity. Columbia Climate School ; Columbia University. 

The London School of Economics and Political Science, & Grantham Research Institute on Climate Change and the Environment. (2025, June 23). New study finds AI could reduce global emissions annually by 3.2 to 5.4 billion tonnes of carbon-dioxide-equivalent by 2035. Grantham Research Institute; Grantham Research Institute.

The ML.ENERGY Leaderboard. (2026, March 17). Ml.energy; GitHub.

UN environment programme. (2022, November 7). How artificial intelligence is helping tackle environmental challenges. UN Environment Programme; UN environment programme. 

United Nations World Food Programme. (n.d.). SKAI. United Nations World Food Programme; World Food Programme. Retrieved March 28, 2026.

United States Environmental Protection Agency. (2026, February 3). Examples of Participatory Science Projects Supported by EPA. United States Environmental Protection Agency; United States Environmental Protection Agency.

University of Florida. (n.d.). Guidance for Researchers. UF Artificial Intelligence; University of Florida. Retrieved April 19, 2026.

Velasquez, M., Andre, C., Shanks, T., J, S., & Meyer, M. J. (2014, August 1). Justice and Fairness. Santa Clara University; Markkula Center for Applied Ethics. 

Velasquez, M., Andre, C., Shanks, T., S.J., & Meyer, M. J. (2014, August 8). Rights. Santa Clara Markkula Center for Applied Ethics; Santa Clara University. 

Yañez-Barnuevo, M. (2025, June 25). Data Centers and Water Consumption | Article | EESI. Environmental and Energy Study Institute; Environmental and Energy Study Institute (EESI).