FORSEE submission to call for evidence of Apply AI Strategy – strengthening the AI continent
Authors: Dr Elizabeth Farries, Dr Alexandros Miniotakis, Dr Patrick Brodie, Prof Eugenia Siapera, Sandra Sieron
AI’s energy intensive infrastructures
It is now widely acknowledged that the development and deployment of artificial intelligence systems come with significant energy footprints and environmental costs.1 Large-scale AI models depend on energy-intensive infrastructures such as data centres, which consume vast amounts of electricity, water, and other natural resources.2 Despite this, public and policy discussions continue to treat AI as largely immaterial, sidelining the physical and ecological realities of digital expansion.3
The regulatory lacunae
While the European Union has expressed a commitment to sustainable AI in various strategic documents4, this recognition has not translated into concrete regulatory mechanisms. The AI Act — the EU’s first comprehensive and legally binding framework for AI governance — acknowledges the environmental impact of AI but its risk classification mostly focuses on AI systems’ technical capabilities and sector of deployment. The EU so far has not adequately integrated the environmental aspect into its policy and risk classification systems. Measures are voluntary and rely on standardisation processes around the consumption of energy by high-risk AI systems.5 This limited articulation of and response to risk largely excludes environmental impact from the regulatory equation, leaving the energy and resource demands of AI systems unaccounted for in policy terms.
Policy consequences of separating AI from its infrastructural scaffolding
One key issue is the artificial separation between AI applications and the infrastructures that support them. Data centres are critical to the functioning of AI systems, yet they fall outside the scope of most AI-specific policy frameworks. This distinction allows environmental risks to be treated as external to AI governance, even though they are structurally inseparable. Data Centres operate in the grey areas between the local, national, and European policy mandates – a form of governance fragmentation leading to novel conflicts over the allocation of land, water, and energy supplies – and the management of the environmental consequences.6 In some cases, where aging infrastructure is already under strain, AI data centers compete with local communities for limited resources like water and energy7; in these cases, public good and ecological care is often sideline in favour of industrial needs.
Moreover, a lack of transparency — often justified through the invocation of trade secrets — prevents regulators and the public from accessing essential information about the energy consumption and environmental footprint of AI infrastructure.
Embedding the environment as a systemic consideration
To align digital policy with the EU’s climate goals and sustainability commitments, environmental impact must be recognised as a systemic risk within AI regulation. This requires an expanded regulatory scope that includes the full AI supply chain, from model training and data storage to ongoing deployment. One avenue is to consider the estimated energy consumption of training AI models in order to determine whether the model proposes a systemic risk. Other options include human rights and ecological impact assessments across the supply chain of sourcing materials, energy, and water. For example, in locations like Ireland where data centres are expanding beyond the capacity of local infrastructure, transition fuels like fossil gas are being deployed to power data centres off the electricity grid8. It must not be left to AI companies to voluntarily self-report these impacts either — independent guardrails must be conceived, assessed, and monitored. Beyond impact and risk assessments, policymakers must acknowledge that digital growth is not limitless and must be subject to environmental boundaries.9
If the EU is to lead in environmentally sustainable AI, its governance frameworks must reflect the real-world impacts of these technologies. Sustainability cannot remain a rhetorical commitment — it must be embedded into the legal and policy structures that shape the future of artificial intelligence.
- Crawford, K. 2021. Atlas of AI: Power, politics, and the planetary costs of artificial intelligence. New
Haven: Yale University Press; Baum, S. D., & Owe, A. (2022). Artificial Intelligence Needs
Environmental Ethics. Ethics, Policy & Environment, 26(1), 139–143.
https://doi-org.ucd.idm.oclc.org/10.1080/21550085.2022.2076538 ↩︎ - Brodie, P. (2023). Data infrastructure studies on an unequal planet. Big Data & Society, 10(1).
https://doi.org/10.1177/20539517231182402 (Original work published 2023) ↩︎ - See for example, member states’ brief and non-prescriptive suggestions to power data centres with
renewable resources ‘where suitable’in their guidance documents Department of Public Expenditure,
NDP Delivery and Reform. (2025, May 7). Guidelines for the responsible use of AI in the public
service. Government of Ireland at p.34
https://assets.gov.ie/static/documents/Guidelines_for_the_Responsible_Use_of_AI_in_the_Public_Se
rvice.pdf ↩︎ - For example, OECD, “Measuring the environmental impacts of artificial intelligence compute and
applications,” 15 November 2022. [Online]. Available: https://www.oecd.org/en/
publications/measuring-the-environmental-impacts-of-artificial-intelligence-compute-and-
Applications_7babf571-en.html; the European Commission’s HLEG’s publication detailing at No6.
Societal and environmental well-being at High-Level Expert Group on AI, “Ethics Guidelines for
Trustworthy AI,” European Commission, 2019. [Online]. Available:
https://digital-strategy.ec.europa.eu/en/library/ethics-guidelines-
Trustworthy-ai; ↩︎ - Warso, Z., & Shrishak, K. (2024, May 21). Hope: The AI Act’s approach to address the environmental
impact of AI. Tech Policy Press.
https://www.techpolicy.press/hope-the-ai-acts-approach-to-address-the-environmental-impact-of
-ai/ ↩︎ - Jansen, F., & Cath, C. (2024, March). Down with data centres: Developing critical policy (CIL#007).
Critical Infrastructure Lab.
https://www.criticalinfralab.net/wp-content/uploads/2024/04/CIL007.pdf ↩︎ - Papaevangelou, C., & Siapera, E. (2025). State, platform capitalism and infrastructural power:
Microsoft’s data centres in Greece 2.0. Platforms & Society, 2, 29768624251323325. ↩︎ - Ireland’s data centres turning to fossil fuels after maxing out country’s electricity grid,
https://www.thejournal.ie/investigates-data-centres-6554698-Nov2024/ ↩︎ - Bashir, N., Donti, P., Cuff, J., Sroka, S., Ilic, M., Sze, V., Delimitrou, C., & Olivetti, E. (2024). The
climate and sustainability implications of generative AI. An MIT Exploration of Generative AI.
https://doi.org/10.21428/e4baedd9.9070dfe7 ↩︎
