The international recognition of the human right to a clean, healthy, and sustainable environment has gained prominence over the past few years, with confirmation of its existence taking place by the United Nations Human Rights Council, the United Nations General Assembly, and most recently the International Court of Justice. This recognition, however, is not entirely new: regional human rights systems in Africa and the Americas have acknowledged this right since the 1980s, and it also exists in many national constitutions. Against this backdrop, a relatively new practice has rapidly gained prominence: the training of Artificial Intelligence (AI) systems. This process, carried out by specialized companies, aims to refine AI systems to deliver increasingly sophisticated outputs. From an environmental perspective, however, studies have linked AI training to significant environmental costs, including water use for cooling operations and high energy consumption, as well as increasing greenhouse gas emissions, given that AI data centers require constant power. This raises a pressing question: as of 2025, does AI training threaten the human right to a clean, healthy, and sustainable environment by contributing to environmental pollution, or does its current scale remain too limited to generate serious human rights concerns?

A Test Case: xAI in the U.S.

Reflections on the evolution of this field can be drawn from the case of xAI, a company dedicated to the development of AI, which, since mid-2024, has operated a data center near Boxtown, Memphis.

Some researchers have found that the presence of pollutants in the air, such as nitrogen dioxide, has increased significantly compared to the pre-xAI period in areas surrounding the data center and, to some extent, in nearby zones. This increase has been linked, with a high level of probability, to the operation of the new xAI data center.

On 17 June 2025, the Southern Environmental Law Center, an environmental advocacy organization in the United States, sent a formal notice to xAI on behalf of the NAACP, the National Association for the Advancement of Colored People, demanding that the company cease operating polluting gas turbines. The notice argued that xAI’s decision “to install and operate dozens of polluting gas turbines without any permits or public oversight is a clear violation of the Clean Air Act”. Boxtown is, in fact, a community predominantly composed of Black residents, originally founded by individuals who were formerly enslaved.

At the local level, the Shelby County Health Department, the public authority responsible for safeguarding community health and environmental standards in the Memphis area, implemented the installation of air monitoring instruments to assess the potential impact of the data center. Similar initiatives were also announced by Memphis Community Against Pollution, which is seeking independent monitoring mechanisms to ensure transparency and accountability.

Yet, despite these measures, residents have continued to raise concerns. Communities in Boxtown and the surrounding areas have reported a range of health issues, from respiratory discomfort to worsening asthma, suggesting a potential link between the operation of xAI and declining health conditions. The World Health Organization has associated similar increases in pollutant levels with the types of symptoms reported by these communities.

The Role of the UN Guiding Principles in the Protection of the Environment and Health

This case is situated within a broader challenge: balancing the rapid expansion of AI training with the protection of the rights to a healthy environment and to health as suggested by the UN Guiding Principles on Business and Human Rights (UNGPs).

According to Principle 13, business enterprises have a responsibility to “avoid causing or contributing to adverse human rights impacts through their own activities, and address such impacts when they occur”. Principle 17 further explains that human rights due diligence should be initiated as early as possible in the development of a new activity, such as the construction and operation of an AI data center, including its management of gas emissions. Moreover, Principle 18 states that this process must involve “meaningful consultation with potentially affected groups”. Applied to the case described above, this would require identifying from the outset which communities may be affected – such as the residents of Boxtown – and taking into account the distinct risks to which they are exposed, including not only environmental damage but also health-related harm.

Finally, Principle 22 emphasizes that “where business enterprises identify that they have caused or contributed to adverse impacts, they should provide for or cooperate in their remediation through legitimate processes”.

Although these principles are not legally binding, they outline a clear framework for action that companies should follow. Since AI training and the operation of data centers are associated with increased greenhouse gas emissions, as well as significant water and energy consumption, establishing a link between AI-related environmental pollution and adverse human rights impacts – whether in the United States or elsewhere – would represent a significant step forward in protecting the human right to a clean and healthy environment, and in preventing future environmental harm. This can contribute to enhancing the protection of human rights.