UN demands environmental transparency from AI companies

What happened?

The United Nations (UN), through its Secretary-General António Guterres, has issued a public call for artificial intelligence companies to transparently disclose the environmental impact of their operations. Specifically, they are required to report on the carbon, water, and land consumption associated with training and deploying their models, as well as that of their data centers and supply chains. The request was amplified this week by Guterres, who urged companies to stop treating the environmental bill as someone else's problem. According to the UN statement, the current lack of transparency prevents governments, investors, and consumers from making informed decisions about the true ecological cost of AI.

Why is it important?

The AI sector is one of the fastest-growing in energy consumption. According to estimates from the International Energy Agency (IEA), data centers could double their electricity demand by 2026, reaching 1,000 TWh annually, equivalent to Japan's consumption. Additionally, the water used for cooling in these centers competes with local communities in water-stressed regions. Until now, most companies have not disclosed these metrics, making accountability and regulatory policy-setting difficult. A study by the University of California, Riverside, estimates that training a large model like GPT-3 consumed 1,287 MWh of electricity and emitted 552 tons of CO₂, equivalent to the lifetime emissions of five average cars. The IEA also notes that data centers already account for 1% to 2% of global electricity consumption, and this figure is expected to skyrocket with the mass adoption of generative AI.

Consequences and reactions

The UN's call is non-binding but increases pressure on giants like Google, Microsoft, OpenAI, and Meta. Some have already begun publishing sustainability reports, but the lack of common standards makes the data incomparable. For example, Microsoft reported a 30% increase in its carbon emissions between 2020 and 2023, largely attributed to the expansion of its data centers for AI. Google, meanwhile, has seen its emissions rise by 48% since 2019. Tech sustainability experts believe this initiative could accelerate the adoption of mandatory metrics, similar to those already in place in the European Union with the Corporate Sustainability Reporting Directive (CSRD). The CSRD, in effect since 2024, requires companies to report their environmental, social, and governance impact, but does not yet specifically cover water and land consumption associated with AI. Organizations like Greenpeace have applauded the measure, though they criticize the UN for not proposing concrete sanctions. On the other hand, the AI Industry Association (AIIA) has expressed willingness to collaborate but warns that publishing detailed data could reveal trade secrets.

What readers should know

Users and companies that use AI services should be aware that each query to a model like GPT-4 or Gemini has a measurable environmental cost. For instance, according to an analysis by startup Hugging Face, a single query to GPT-4 can consume approximately 0.4 kWh of energy, enough to charge a smartphone for a month. On a global scale, daily queries to generative AI models could require the energy of a small nuclear power plant by 2027. The transparency demanded by the UN would allow consumers to make informed decisions and regulators to design incentives for energy efficiency. It could also drive the adoption of renewable energy sources in data centers. Companies like Google have already committed to operating on 24/7 carbon-free energy by 2030, while Microsoft plans to be carbon negative by 2030. However, without standardized metrics, it is difficult to verify these commitments.

Historical context

This is not the first time the UN has intervened in tech issues. In 2021, the organization had already warned about the digital divide and the risks of AI to human rights. However, this is the first time it has specifically focused on environmental impact, reflecting growing concern over climate change. The initiative is framed within the Sustainable Development Goals (SDGs), particularly SDG 13 (Climate Action) and SDG 12 (Responsible Consumption and Production). Additionally, the UN has established the High-Level Advisory Body on Artificial Intelligence, which in its 2023 report recommended creating an intergovernmental panel on AI similar to the IPCC. This call for environmental transparency could be the first step toward a global regulatory framework. Comparatively, the pressure on the tech industry recalls UN efforts in the 1990s to get tobacco companies to disclose health risks, or more recently, demands for transparency on personal data use after the Cambridge Analytica scandal.

In summary, the UN's call marks a milestone in AI governance by focusing on its ecological footprint. While lacking legal force, it lays the groundwork for future regulations and increases the responsibility of tech companies. The challenge now is to translate this pressure into concrete, verifiable standards that allow measuring and reducing the environmental impact of an exponentially growing industry.