Sustainability is among the most contested topics in contemporary debates around AI. While the industry’s meteoric growth has sparked growing public concern about its environmental impacts, Big Tech companies have gone to great lengths to portray AI as not just sustainable but positively good for the environment. Independent research suggests many such claims are unsubstantiated or outright misleading.

One underappreciated reason this matters is that most of the Big Tech companies that dominate the AI industry – like Microsoft, Amazon, Meta and Google – are also consumer-facing companies. Misleading practices related to consumer transactions are illegal under EU law, and the 2024 Directive on Empowering Consumers for the Green Transition has just introduced stricter rules on greenwashing. This blog post argues that some widespread practices in the AI industry – like making vague and unsubstantiated claims about sustainability, claiming carbon neutrality targets while increasing emissions, and developing AI for use in fossil fuel production – could in some circumstances give claimants a strong case for violations of EU consumer protection law.

Corporate “sustainable AI” messaging

The AI boom has brought newfound public attention to the environmental impacts of digital infrastructure. Data centres’ energy and water use have attracted particular attention, due to the scale and speed of construction, which is impacting local communities’ access to water and electricity, as well as aggregate emissions. But data centres are just the tip of the iceberg. Researchers have mapped diverse environmental harms throughout global AI value chains. These include emissions, water use and pollution from semiconductor manufacturing; consumption of minerals like copper and rare earths (which doesn’t just cause direct environmental harms through mining, but also diverts scarce resources away from decarbonisation); increased e-waste; and the use of AI to facilitate harmful activities, like fossil fuel production.

Big Tech companies have worked hard to allay such concerns. Amazon, Apple, Google, Microsoft and Meta all have longstanding sustainability policies and carbon neutrality targets, and are major purchasers of carbon offsets and renewable power purchase agreements (though some have recently watered down these commitments). They also systematically conceal key information about environmental impacts, while promoting the narrative that AI will enable long-term, economy-wide sustainability improvements which outweigh its immediate environmental costs. Critics have shown that such purported benefits remain largely speculative, and these companies’ sustainability initiatives are far from sufficient to counter their rapidly-growing fossil fuel consumption. This raises the question: what does EU law say about these kinds of claims?

How does EU law define greenwashing?

High-profile greenwashing cases have previously been brought under the 2005 Unfair Commercial Practices Directive (UCPD), which bans misleading acts (Article 6) or omissions (Article 7) likely to affect the transactional decisions of an average consumer. However, the 2024 Directive on Empowering Consumers for the Green Transition has just introduced more specific rules on greenwashing. It adds three additional categories to Annex I UCPD, which lists “commercial practices which are in all circumstances considered unfair”:

• making “generic environmental claims” (e.g. “environmentally friendly”, “eco-friendly” or “green”) without being “able to demonstrate recognised excellent environmental performance relevant to the claim”
• making environmental claims about an entire product or business which actually concern only part of the product or the business’ activities
• claiming a product is carbon neutral, or has reduced emissions or positive environmental impacts, based on offsets

Many such claims could already have been challenged as misleading commercial practices. However, this requires showing that they would affect the decisions of an average consumer. Where the new rules apply, this is no longer necessary: practices listed in Annex I are always banned. This should also lead to a more consistent EU-wide approach. Member states were required to transpose the 2024 Directive by March 2026 and it should become fully applicable by September 2026, but it is already being invoked in litigation. Interestingly, it has been used not just to challenge the specific practices that are now banned, but also to support the general principle that consumer interests encompass environmental concerns and greenwashing should be penalised, informing the interpretation of the more general ban on misleading practices.

The Commission also proposed a second Green Claims Directive in 2023, which would require all environmental claims to be scientifically substantiated and independently verified. Whether and in what form this proposal will ultimately be passed remains uncertain, so below we focus on the existing UCPD rules, as amended by the 2024 Directive.

When are sustainability claims consumer-facing practices?

A key precondition for any of these rules to apply is that the UCPD only regulates “business-to-consumer commercial practices” which are “directly connected with the promotion, sale or supply of a product to consumers”. Clearly, this could only apply to consumer-facing tech companies – but as noted above, that includes most of the Big Tech companies which dominate the AI industry. “Products” in the UCPD also include digital services and content.

However, it is not enough that a company is generally bound by the UCPD: the greenwashing rules only apply where the specific practice at issue is “directly connected” to the promotion, sale or supply of consumer products. Under what circumstances would that apply to AI companies’ claims about sustainability?

The UCPD does not apply to investor-facing communications or mandatory regulatory disclosures, like corporate sustainability reports. Other corporate communications, like blog posts and sustainability policies, are a grey area: they are publicly available, but generally target expert audiences more than consumers. Recital 7 UCPD states that it excludes “commercial practices carried out primarily for other purposes” than influencing consumer transactions, like “annual reports and corporate promotional literature”. However, Anna Beckers argues that sustainability policies and other “promotional literature” often do aim to influence consumer behaviour, by shaping public perceptions of a company, so they should be included unless they are not addressed to consumers at all.

This interpretation is broadly supported by a recent French ruling that TotalEnergies’ claims about its net zero “ambitions” were misleading consumers (because it was continuing to pursue new oil and gas developments, which the IPCC has established is incompatible with the Paris targets). The court held that press statements and posts on Total’s general corporate social media channels were not related to consumer transactions, but claims on its website were adjacent to information about consumer-facing products and aimed to influence consumer decisions. Similarly, where Big Tech companies promote claims about AI and sustainability through public-facing channels like mass advertising or consumer-facing websites, it is strongly arguable that these claims are intended to influence consumers’ decisions to use their products, and are therefore subject to the UCPD

What does this mean for “sustainable AI” claims?

Determining whether specific claims are unfair or misleading would, again, be quite fact-specific (and we hope to explore this further in future research). However, a brief survey of some existing research on Big Tech companies’ marketing and PR messaging around sustainability points to some widespread practices which could fall foul of these rules, and warrant further investigation.

Clean energy & carbon neutrality

Big Tech companies widely promote and publicly advertise their net zero/carbon neutrality targets. Given their substantial (and growing) fossil fuel consumption and direct emissions, these claims rely on carbon offsets and renewable power purchasing agreements. Importantly, the 2024 Directive’s per se prohibition of claiming carbon neutrality based on offsets only applies to claims about specific products – not claims about a business as a whole. However, as the Commission’s guidance emphasises, claims that a business is carbon neutral are still subject to the generally-applicable ban on misleading practices likely to affect the decisions of an average consumer.

Research on corporate offsetting schemes generally, and Big Tech specifically, has called into question these schemes’ integrity (e.g. the extent to which they actually make additional offsetting or renewables projects happen, or merely take credit for existing projects). More fundamentally, the idea that corporate emissions can be “offset” is questionable, since the Paris targets require ‘every feasible emission cut that can be achieved anywhere’. In practice, adding renewable capacity has historically covered increasing aggregate energy use, rather than decreasing fossil fuel consumption.

In light of this research, companies relying on renewable power investments and offsetting schemes to support “net zero” claims already appear vulnerable to allegations of misleading practices. Here, such allegations would be bolstered by the fact that Big Tech companies’ emissions are not just substantial but dramatically increasing, due to their vast investments into highly computing-intensive approaches to AI development. An average consumer would probably assume that a company claiming to be committed to net zero is “heading towards the goal, not away from it”.

Importantly, the UCPD bans misleading omissions as well as explicit statements – including where “omitted” information is technically available, but not clearly brought to consumers’ attention (for example, where it is not in the same advert as the potentially-misleading communication). Thus, where companies selectively present accurate information (e.g. that they are investing in renewable energy) but omit other relevant information (e.g. that they are also sharply increasing their fossil fuel consumption) this could also be regarded as misleading consumers, by falsely suggesting that they are shifting away from fossil fuels and towards clean energy.

Vague “AI for sustainability” claims

Energy and AI researcher Ketan Joshi recently compiled a database of claims about how AI could promote sustainability (mostly from independent sources, but often based on company blogs or corporate research). He found that most were unsubstantiated, overtly misleading and/or excessively vague: for example, conflating relatively small-scale and efficient “AI for sustainability” tools with resource-intensive generative AI tools which have little to no environmental benefit.

If such claims are made in a context with a direct connection to the sale or promotion of consumer products, there is a strong case that they are misleading under the UCPD. Relevantly, a recent Dutch case against airline KLM held that claims could be misleading because they were vague and unsubstantiated, even though not strictly untrue. Vague claims that AI is good for sustainability or has a “positive impact on the planet” could also violate the per se bans on unsubstantiated generic environmental claims and on broad claims which are only true of part of a company’s activities.

AI for unsustainability?

Something else Big Tech companies tend to omit from their public-facing messaging about “AI for sustainability” is that AI is also widely used to optimise and scale up environmentally harmful activities. In the most notable example of what may be called “AI for unsustainability”, Microsoft, Google and Amazon all do significant business with fossil fuel companies and develop specialised AI services for oil and gas production. As the Enabled Emissions campaign has highlighted, these projects’ climate impacts could dwarf companies’ offsetting and renewables initiatives – not only by increasing production from existing reserves, but also by helping locate new ones, and enabling efficiencies which make more projects economically viable.

Business-to-business transactions and marketing are not directly regulated by the UCPD. However, these activities could be relevant in establishing that Big Tech companies’ consumer-facing sustainability claims are misleading – for example, that they are omitting key information when they claim that AI is being used to accelerate the green transition, or that their net zero targets are misleading. The Total case set a particularly important precedent by holding that it is misleading for a company to claim to be committed to net zero while pursuing new oil and gas projects.

Enforcement & future outlook

Greenwashing has historically been one of the most common legal bases for corporate climate litigation. This brief analysis suggests that leveraging EU consumer protection law could also be a promising legal route to challenge unsustainable practices in the AI industry. At a time when governments in Europe and elsewhere are generally committed to promoting AI development, deployment and infrastructure construction, and sometimes actively undermining accountability for environmental impacts, greenwashing lawsuits offer a route to contest companies’ environmental impacts which is not dependent on regulatory action.

Expectations should be realistic, however. The UCPD only applies to consumer-facing statements. That excludes more subtle and pervasive ways that Big Tech companies shape public debates around AI and sustainability, like lobbying and research funding. The ease of enforcement will also depend significantly on national procedural frameworks. Most importantly, the remedies available are limited. Even successful greenwashing claims cannot restrict companies’ actual unsustainable business practices, but only their consumer-facing communications – typically leading to relatively small damages and/or orders to withdraw misleading claims.

Nonetheless, litigation can have impacts beyond these immediate legal consequences: for example, attracting media coverage and damaging companies’ reputations (one recent study found that media accusations of greenwashing measurably affect companies’ stock prices). It could help publicise and politicise the environmental harms of AI, including underappreciated issues like “enabled emissions”. The strenuous efforts Big Tech companies make to present themselves and their AI tools as sustainable suggest that they would see this as a meaningful cost.