Green promises, grey realities

Artificial intelligence is often presented as a technology of solutions. Tech companies, investors and policymakers often claim AI will make industries more efficient, accelerate scientific discovery, optimize energy systems, and perhaps even help address the climate crisis. But what happens when the story of “sustainable AI” becomes part of the problem?

This question is at the center of a new contribution by Rachel Griffin, Postdoctoral Researcher in Prof. Jat Singh’s Compliant and Accountable Systems group at RC Trust, and Barrie Sander, Assistant Professor of International Law at Leiden University. In their blog post Contesting Big Tech’s AI Greenwashing: Assessing ‘Sustainable AI’ Claims Under EU Consumer Protection Law, published on Verfassungsblog, they examine whether some sustainability claims made by major technology companies could be challenged under EU consumer protection law.

The starting point is a growing tension in public debates about AI. On the one hand, large technology companies increasingly promote AI as environmentally beneficial: a tool for optimization, innovation, and long-term sustainability. On the other hand, independent research points to the substantial environmental costs of AI infrastructures. Data centers require vast amounts of energy and water. Semiconductor production consumes resources and causes pollution. AI hardware contributes to electronic waste. And AI systems are also used in areas that can directly support environmentally harmful activities, including fossil fuel exploration and production. Like the global impacts of climate change, the local environmental impacts of AI infrastructure and supply chains disproportionately affect disadvantaged communities and poorer countries. 

Griffin and Sander do not simply ask whether AI is sustainable or unsustainable. Their focus is more precise: how do powerful technology companies talk about AI and sustainability, and when might this communication become misleading?

This matters because sustainability claims  shape how consumers, policymakers, investors, journalists, and the wider public understand the role of AI in society. If a company presents itself as moving towards carbon neutrality while its emissions continue to rise, or if it promotes vague claims about “AI for sustainability” without clear evidence, such communication can influence public trust and consumer choices. In that sense, greenwashing is not only a reputational issue. It is also a governance problem.

The legal angle of the article is particularly timely. EU consumer protection law already prohibits misleading commercial practices, and recent updates to European legislation have introduced stricter rules on greenwashing. Griffin and Sander argue that some common forms of “sustainable AI” messaging could, under certain circumstances, fall within this legal framework. This may apply, for example, to vague and unsubstantiated environmental claims, broad sustainability statements that only relate to part of a company’s activities (for example, when companies make broad claims that AI will bring significant sustainability benefits, based on a few unrepresentative cases), or carbon neutrality claims based on offsetting.

The blog post carefully avoids suggesting that consumer protection law can solve the environmental problems of AI by itself. The scope of these greenwashing rules is limited: they apply only to consumer-facing communication, not to all corporate lobbying, investor messaging, research funding, or business-to-business activities. Even successful legal challenges would only address misleading claims rather than directly restrict environmentally harmful business practices.

Yet this limitation is part of the insight. Consumer protection law may not stop the expansion of energy-intensive AI infrastructures. But it can offer a concrete tool for scrutinizing how these infrastructures are presented to the public. It can help ask whether people are being given a fair picture of the environmental trade-offs behind AI products and services. And it can support further public debate about the environmental costs of AI by making contested sustainability narratives more visible.

This is where the contribution connects strongly with RC Trust’s research agenda. Trustworthy technology is not only a matter of technical reliability, privacy, or security. It also depends on whether institutions and companies can be held accountable for the claims they make about their technologies. If AI systems are promoted as beneficial, green, or climate-friendly, those claims need to be open to scrutiny.

Rachel Griffin’s work at RC Trust examines how digital technologies are governed, how risks are defined, and how regulation can better account for power, accountability, and public trust. In Jat Singh’s Compliant and Accountable Systems group, this perspective fits into a broader interdisciplinary approach at the intersection of computer science, law, and society. The group asks how technologies can be designed, deployed, and governed in ways that respond to legal, regulatory, and digital rights concerns.

Published on Verfassungsblog, an open-access forum for legal and political debate at the interface of academia and society, the blog post contributes to an urgent conversation. As AI becomes more deeply embedded in economic and public life, its environmental consequences will not only be measured in data centers, emissions, or water use. They will also be shaped by the stories companies tell about them.

Griffin and Sander’s contribution helps sharpen the tools for reading those stories critically and questioning whether companies’ claims can withstand legal, scientific, and public scrutiny.