The UN calls on AI companies to be transparent about their environmental impacts.

      The United Nations is urging artificial intelligence companies to cease regarding the environmental impact as someone else's concern. In a statement amplified this week by Secretary-General Antonio Guterres, the organization is calling on the companies driving the AI surge to reveal the carbon, water, and land usage their systems entail, and to transition their data centers and supply chains to renewable energy sources before the financial burden falls on the communities least equipped to handle it.

      This demand for disclosure is backed by new evidence. A report from the UN University Institute for Water, Environment and Health (UNU-INWEH) from June 2026 warns that the environmental impact of AI is rapidly increasing across all three categories, and it urges immediate action to keep the technology within what it describes as planetary boundaries.

      This report shifts the discussion, which has predominantly focused on electricity, to also address water and land usage. The significant figure is electricity consumption. The report predicts that data centers could consume approximately 945 terawatt-hours of electricity annually by 2030, a figure comparable to the total annual electrical consumption of Pakistan, Bangladesh, and Nigeria, three countries with a combined population of over half a billion.

      On this trajectory, the energy requirements for AI could be nearly three times what those three nations consume collectively. The demand for water is a less obvious yet significant factor. The cooling of servers and the energy needed to operate them contribute to what the authors refer to as AI’s water footprint, with projections indicating that water usage related to AI could equal the basic domestic water needs of 1.3 billion people by the end of the decade.

      Land use, which is primarily related to the space and infrastructure needed for power generation, is the third component the report includes in a traditionally singular narrative. None of this information is framed as a reason to abandon AI; instead, it advocates for a different approach to its development. The report proposes a responsible AI ecosystem based on transparency, efficiency by design, equity, lifecycle accountability, global collaboration, and sustainable practices.

      Transparency, the first principle, is especially highlighted by Guterres since nearly all other aspects rely on companies being truthful about their actual resource consumption. Currently, there is a lack of such honesty, which is a significant issue. Operators seldom share detailed data about the water usage of specific data centers or the emissions associated with particular models, leaving experts to make estimates.

      This push for transparency aims to transform speculative assessments into verifiable metrics that regulators and the public can audit. The UN emphasizes that these costs should not be quietly transferred to vulnerable communities, which are often the locations chosen for data centers due to inexpensive land and lenient regulation.

      This warning comes as AI infrastructure is already being developed at an unprecedented scale. U.S. utilities are preparing for a $1.4 trillion capital influx to support data centers projected by 2030, and initiatives like Meta’s gas-powered Hyperion campus in Louisiana illustrate how energy shortfalls are being addressed through fossil fuels rather than the clean energy alternatives the UN advocates.

      Regulators have begun to take notice, with the EU instructing major tech companies to align their data centers with climate objectives, while numerous startups are exploring methods to reduce energy consumption. However, the report stops short of imposing strict limits or naming companies, leaving the more challenging political enforcement to governments.

      The report’s contribution lies in providing a framework and relevant data, offering a way to measure an industry that has grown more rapidly than the means to monitor it. The complete UNU-INWEH report is accessible from the institute, but the actual disclosure still depends on the companies involved.