Despite its increasing role in tackling climate change and fostering economic development, artificial intelligence (AI) is criticized for the environmental impacts it causes: high electricity consumption, rising carbon emissions, and increasing water use. However, regulatory frameworks across the world are shifting towards mitigating this issue, albeit extremely slowly.
See also: The impact of artificial intelligence on the environment | Experts’ Opinions
The much-needed regulatory updates are mainly on the radar of developed countries and regions, such as the European Union and the United States. Their focus is on ensuring that the expansion of AI does not impede environmental sustainability goals, while the mechanisms involved will feature new sustainability provisions in current or future AI legislation and principles that emphasize eco-friendly AI utilization.
AI’s impact on the environment
The growing popularity of AI models is triggering increased energy consumption, which leads to a global increase in CO2 emissions, that are known to be one of the reasons for rising temperatures across the globe. Thus, increased usage of AI caused a 72% increase in data center energy use from 2019 to 2023, with higher CO2 emissions and water consumption as AI data centers depend on water-intensive cooling processes. According to Arbor, each interaction with an AI model, such as ChatGPT, leads to roughly 4 grams of CO2 being emitted into the atmosphere. This means 14,000 tons of CO2 daily, and roughly 5.1 million tons annually. Considering the increasing use of AI models and more complicated prompts and queries, this figure is only likely to grow.
How is AI regulated from an environmental perspective?
European Union
- The EU is among the first regions to regulate AI, with its EU AI Act being approved on August 1, 2024, but it is due to be fully applicable in August 2026.
- In addressing the environmental impacts, the Act incorporates provisions including:
-
- Article 40 requires standardization institutions to provide regulations for reducing energy and other resource consumption for high-risk AI technologies.
- Regulatory sandboxes emphasizing the controlled testing of AI technologies while protecting the environment, biodiversity, and climate change.
- A reference to the EU commitments to monitoring the energy efficiency of data centers and eco-design requirements for servers and computers.
- Transparency regulations for General Purpose AI (GPAI) models that require developers to monitor and document the models’ environmental footprint.
-
- Weaknesses: The Act’s environmental provisions, relying on standardization bodies, is delaying progress and limiting impact since there are profit-driven members who will resist costly but greener alternatives. Also, the adherence to the Act’s standards on environmental sustainability is voluntary and not compulsorily publicized, limiting positive results.
The United States of America
- Sectoral Plan: Currently, the U.S. does not have detailed AI-oriented environmental regulations in force. However, the U.S. Environmental Protection Agency (EPA) is leveraging AI to ensure adherence to environmental policies such as the Clean Air and Clean Water Acts. AI is enhancing the timely monitoring and enforcement of the regulations.
- NIST Framework: The National Institute of Standards and Technology (NIST) is establishing sustainability frameworks and evaluating AI systems by identifying performance indicators such as carbon footprint and energy consumption.
United Kingdom
Principles-Based System: The UK’s regulatory framework is based on principles such as transparency, safety, and accountability, with the main function focusing on assessing AI risks. Environmental sustainability is one of the fundamental principles, but there are no mandatory environmental regulations specific to AI.
Australia
Compulsory Safeguards: Australia requires disclosures, auditing, and accountability for high-risk AI, with its National Artificial Intelligence Ethics Structure incorporating environmental protection as a fundamental principle.
Canada
Artificial Intelligence and Data Act( AIDA): This is expected to reduce the impact of high-performance AI technologies by enforcing restrictive measures to mitigate harm and promote transparency. Ecological protection is not yet fundamental, but may be considered as the law changes.
Japan, India, and South Korea
- Japan: Principles emphasize human-centric, data-protection, fairness, and safety, but do not mandate environmental sustainability.
- India: Currently lacks specific regulations for AI, but tenets of responsible AI encompass equality, accountability, reliability, and openness.
- South Korea: Policies emphasize sustainable AI research and environmental sustainability, but there are no mandatory environmental AI regulations.
Environmental concerns over increasing AI use
Despite AI features rapidly transforming almost every sector, its energy consumption rate is raising environmental concerns. If a user logs in to a feature like ChatGPT, a process consumes at least 10 times as much electrical energy compared to a standard Google search. Also, the International Monetary Fund (IMF) 2025 report indicates that AI could triple the world’s electricity usage by 2030, equating to the yearly energy consumption of entire countries such as India, and affecting environmental sustainability.
See also: Power-hungry, water-thirsty: How AI is quietly devastating the planet
Moreover, for daily user interactions with Large Language Models such as GPTs, a different issue emerges. If unregulated, the use of AI is expected to significantly increase greenhouse gas emissions due to the consumption of large amounts of electricity for data processing in data centers and model training.
In the U.S., home to the majority of data centers in the world, the IMF estimates that AI could cause an increase in electricity prices and emissions by at least 8.5% and 5.5%, respectively. Worldwide, the climate burden could surge by 1.2% due to AI-related emissions.
Distribution of data centers
The AI data center market is expanding at an annual growth rate of approximately 28%, surpassing conventional data centers. Developed nations are leading with most of the AI-related facilities due to reliable technological infrastructure, abundant energy sources, and investment in the digital economy. The U.S. dominates in the number of data centers and AI investment, with at least 5,381 data centers and almost US$55 billion in funding. Germany is the second highest with around 522 data centers and about US$2 billion investment in AI.
Figure 1: Top 10 Nations with the most data centers as of March 2024
Source: Statista
Wrap Up
AI offers immense opportunities for innovation and economic development, but there is a need to regulate its environmental impacts. Developed nations are at the forefront in formulating AI environmental regulations. However, these countries are also responsible for the increased use of energy and other resources due to extensive data center development. On the other side of the coin, developing countries are struggling to draft and implement such regulations due to the lack of AI data stations and inadequate resources to reduce AI’s ecological footprint. Therefore, in mitigating the environmental impact of AI, nations should focus on implementing regulations that mandate the use of renewable energy sources in data centers.
