Why is enforcement of the DSA systemic risk framework still ignoring environmental risks?
By Rachel Griffin
This post argues that environmental risks stemming from major online platforms—including those linked to digital infrastructure, generative AI, targeted advertising and e-commerce—straightforwardly fall within the scope of platforms’ risk management obligations under Articles 34 and 35 of the DSA. The author questions why neither platforms nor the European Commission have meaningfully addressed these risks, and argues that their continued omission reflects political priorities more than legal constraints.
Since the start of the current AI boom, there has been growing public concern about the environmental impacts of digital infrastructure. Even though the specialised hardware and vast data centres used for large AI models are particularly resource-intensive, we should remember that older digital services we now see as unremarkable – like online platforms and streaming services – also rely on material infrastructure with significant environmental impacts. There is also substantial overlap between the ‘big tech’ companies dominating the AI market and those operating leading consumer platforms – many of which are rapidly integrating highly energy- and resource-intensive generative AI systems into their platform services.
This raises the question: why are environmental impacts are still being ignored in the enforcement of the DSA? Legal academics increasingly agree that the DSA’s systemic risk management obligations could extend to environmental risks. But so far, this possibility is nowhere to be seen in enforcement and compliance practices. The European Board for Digital Services recently published its second annual report summing up how platforms have identified and addressed systemic risks. As in the first report, environmental impacts of their business operations are entirely unmentioned. Nor has the Commission addressed these issues in any of its 14 enforcement proceedings or 63 requests for information. As Europe reels from heatwaves that have killed thousands of people, while platform companies’ emissions shoot upwards, why are systemic risks to the environment not on the agenda?
Why we should worry about platforms’ environmental impacts
Just like using an AI chatbot, scrolling videos on TikTok relies on a sprawling global network of data centres, telecommunications infrastructure and commodity supply chains, which have significant environmental impacts. Remarkably, a 2024 study estimated that just maintaining then-current levels of online content consumption (including web browsing, video and music streaming, video calls and social media) would account for about 40% of the world’s per capita carbon budget compatible with the target of 1.5 °C of global heating, as well as around 55% of sustainable minerals and metals consumption. Since 2024, some leading platforms have rushed to incorporate AI video generation tools into their advertiser services (as well as promoting AI-generated content in ‘organic’ user feeds). Generating videos is one of the most energy- and resource-intensive AI tasks, so it’s almost certainly significantly driving up online platforms’ contribution to these environmental impacts.
Many platforms – especially but not only social media – rely heavily if not primarily on targeted advertising for revenue. This raises several further environmental concerns. First, advertising exacerbates the impacts of digital content consumption described above, because it incentivises companies to design their platforms to maximise ‘engagement’. That means platforms are geared towards encouraging users to spend as much time consuming content as possible – often particularly data- and energy-intensive forms of content like short videos (apparently, the average British TikTok user ‘dedicates more than 42 hours a month to watching videos which rarely last longer than thirty seconds’). Running advertising systems themselves – profiling users, generating audiences and distributing ads – also requires processing and transmitting even more data, beyond what would otherwise be necessary for platforms to function.
To the direct impacts of digital infrastructure, we can add the environmental impacts of products sold via targeted advertising and e-commerce. Obviously, advertising systems are designed to induce users to purchase more goods and services – regardless of whether they need them. In an ongoing project on environmentally harmful uses of AI, Barrie Sander and I argue that this is an inherently unsustainable business model, because current levels of material consumption in wealthy regions like Europe urgently need to decrease in order to enable basic living standards for everyone in the world within the physical limits of Earth systems. Similarly, the basic business model of e-commerce platforms is to encourage consumers to purchase as much as possible – often by promoting cheap, disposable products transported from the other side of the world, via resource-intensive and polluting global logistics networks.
Finally, as key intermediaries governing people’s access to news, information and social interaction, platforms also influence climate politics. Online advertising is an important channel for anti-environmentalist interest groups to disseminate both overtly anti-environmental policy messages, and more subtle forms of ‘greenwashing’ or tech solutionism. Algorithmic recommendation systems and revenue redistribution programmes designed to incentivise the production of ‘engaging’ content may also boost the visibility of climate dis- and misinformation.
Are environmental impacts a systemic risk?
As I’ve argued in my research on the DSA, there’s no objectively correct answer to the question of whether something is a risk (much less a systemic one). Risks aren’t objectively-existing external situations – they’re a way of framing issues and making decisions. What issues we identify as risks and how we understand them depends on what we value and what we want to achieve (and who gets to make those decisions). However, in my view, it’s certainly the case that the DSA systemic risk framework can be interpreted as including environmental impacts. In that sense, the inattention to these issues by the Commission and major platforms says a lot about what they do and don’t value.
I argued in 2023 that the DSA’s systemic risk provisions could extend to environmental impacts discussed above. Since then, a number of other legal scholars have come to the same conclusion. First, as Europe’s recent heatwaves should have reminded us, climate change and other environmental crises significantly and directly impact several of the ‘systemic risk’ areas that large platforms are required to address under Article 34(1) DSA – notably public health, public security, physical and mental wellbeing, and fundamental rights. Second, Article 34(1) also specifically refers to risks ‘stemming from the design or functioning of [a VLOP’s] service and its related systems, including algorithmic systems, or from the use made of their services’ (emphasis added). This indicates that risk assessments are not just meant to cover risks associated with platforms’ technical features and content policies, but must consider how these services are used in their social and economic context. Finally, some scholars have argued that Articles 34-35 should be interpreted purposively in light of EU primary law and fundamental rights law, to promote a high level of environmental protection. I agree with this, but I would also stress that including environmental risks doesn’t require a strained or artificial reading of the DSA – it fits quite straightforwardly within a literal reading of the legal text.
Against this view, leading DSA expert Martin Husovec has argued that the DSA’s stated aim is ‘a safe, predictable and trusted online environment’, and environmental impacts lack a sufficient ‘link to online environment’. However, this seems at odds with the explicit inclusion in Article 34(1) of other systemic risk areas, like ‘public health’ and ‘electoral processes’, which can only sensibly be interpreted as involving platforms’ indirect impacts on broader ‘offline’ social systems. In any case, this sharp distinction between on- and offline environments is questionable. ‘Online environments’ are not immaterial parallel dimensions – they are energy-, resource- and labour-intensive material infrastructures that are situated within, impact and are impacted by wider environmental systems.
At a minimum, including environmental risks in systemic risk assessments should strengthen transparency. Platforms should be detailing the direct and (estimated) indirect environmental impacts of their services in their annual reports. This could draw from and might partially duplicate existing corporate sustainability reporting, but should also be broader in scope. As mentioned above, the DSA systemic risk provisions are broadly framed and cover indirect societal impacts associated with the use of services, which aren’t always included in existing corporate sustainability reporting protocols.
More practically, legal scholarship characterising environmental issues as systemic risks has suggested several relevant mitigation measures (of varying relevance for different platforms, depending on their features and business models). Companies could take energy efficiency, sufficiency and ‘sustainability by design’ measures to reduce the energy footprint of their digital infrastructure. The DSA also requires a dedicated risk assessment when platforms launch new products and features; where this assessment indicates excessive environmental impacts (as for example with ubiquitous AI video generation) companies should simply avoid launching such features. They could radically reduce the user data collected for advertising purposes – which could also help address various other societal risks associated with targeted advertising. They could refuse to advertise, promote or distribute particularly environmentally damaging products and services, and could take more robust and targeted measures to crack down on environmental disinformation and greenwashing (which will often qualify as illegal content under EU consumer protection law, meaning it’s undoubtedly within the scope of the systemic risk framework).
A gap in regulatory enforcement
On a practical level, though, this growing academic consensus around the inclusion of environmental impacts of systemic risks seems to have had essentially no impact on regulatory enforcement and compliance practices. This could in itself be seen as problematic, given that the DSA theoretically attaches great importance to multistakeholder participation and consultation. Recent years have seen an explosion of research and public concern about the environmental impacts of digital infrastructure, as well as a growing body of research on the relevance of environmental risks in the DSA. Recital 90 explicitly states that platforms’ risk assessment should consider ‘the best available information and scientific insights’.
Yet VLOPs’ compliance reports so far have essentially ignored environmental impacts. Related issues are sometimes mentioned in passing (e.g. as regards climate misinformation, or whether marketplace sellers disclose mandatory environmental information about products). However, the environmental impacts of platforms’ own business operations go entirely ignored. That’s unfortunate but not surprising: VLOPs have very little incentive to voluntarily draw attention to harmful aspects of their business practices by discussing risk areas other than those explicitly mentioned in Article 34.
What’s more problematic is that the Commission has ignored environmental risks in DSA enforcement. The legal-institutional context can also help explain this. Since any enforcement actions can be legally contested by well-resourced corporate defendants, the Commission is generally incentivised to focus on ‘winnable’ cases involving risks explicitly mentioned in Article 34, like child safety or illegal content. Yet coercive regulatory action would not be the only option available here. The Commission has not even requested any information from platforms about environmental risks, the easiest and least intrusive measure it could take. It could also easily issue non-binding guidance that recommends covering environmental risks. Julian Morgan has recently argued that the possibility to issue such guidance under Article 35(3) has more generally been under-used, as the Commission seems to have favoured opaque negotiations with platform companies over transparent and accountable public deliberation.
Overall, then, the failure to address platforms’ environmental issues within the DSA systemic risk framework looks less like a legal constraint and more like a political choice – one that’s in keeping with the current Commission’s overall deregulatory and anti-environmentalist agenda. But as climate breakdown continues to escalate, and the tech sector’s role in exacerbating it continues to grow, the failure to even acknowledge platforms’ environmental impacts in DSA enforcement is becoming harder to justify.