ľ�ϸ���Ӱ��

(by Alice Pederzoli, student in the master programme European Law – Utrecht ľ�ϸ���Ӱ��, 3 March 2025)

The AI Action Summit 2025: from Collaboration to Competition

On 10 and 11 February 2025, heads of state, tech giants’ CEOs, and scientists arrived in Paris from all over the world to take part in the conversation on the future of AI at the Artificial Intelligence Action Summit. As French President Emmanuel Macron told CNN, Europe is “” on AI investments, thereby making the goal of the EU clear from the outset: to become a frontrunner in the race to AI supremacy.

The focus of the Summit was supposed to be on global cooperation for scientific solutions to the progress of AI in respect of the . However, EU Commission’s President, Ursula von der Leyen, was eager to announce that the Union will be investing €50 billion in the investAI initiative, €8 billion more into AI Factories, and €20 billion in AI Gigafactories, fuelling the competition as she declares that

The abrupt shift of focus of the Paris Summit sparked several debates as it became increasingly evident that the EU might not be so much “lagging behind” on AI investments but more on sustainability goals. The environmental impact of AI, in fact, was overlooked in favor of a race for technological dominance, which raises concerns for the long-term effects of resource-intensive AI infrastructures.

The Environmental Cost of AI

The disappointment towards the outcome of the Summit is widespread, with many experts considering it a because of the well-known effects of AI technologies on the environment. ChatGPT-3, for example, requires approximately to answer 10 to 50 questions. Similarly alarming is the carbon footprint of training an AI model such as ChatGPT-2, which is around . Importantly, the more advanced the model, the more water required and the more CO2 released to the environment.

These and many others are the reasons that brought professor and author John Naughton to tell The Guardian that “,” while other experts went so far as to state that . Evidently, the issue is real and urgent. However, while raising awareness about the situation is essential, is merely acknowledging the issue enough to drive change?

If the EU truly wants to lead in AI, it cannot afford to treat sustainability as an afterthought. Now that we’ve seen that investments are not lacking, the real question is how the EU can ensure that these funds are directed towards developing future-proof, sustainable AI that aligns with its long-term environmental commitments rather than undermining them.

The Role of AI in the European Green Deal…

It was only when AI models such as ChatGPT were put in the hands of the general population — in late 2022 — that most of us began focusing on the impacts that these technologies would have on our quality of life, our jobs and our planet. Institutions all over the world, however, had their eye on AI way before then, hoping to develop models that could be applied to all fields.

The EU was no different, and in 2018, it had already set up the to develop AI in line with the EU’s shared values as enshrined in the Treaties. Already then, sustainability was seen as one of the main long-term challenges to tackle in relation to AI.

With this goal in mind, AI was given an important role in the European Green Deal already in the European Commission’s dated February 2020. Not only did it argue that AI was going to become a useful asset for the achievement of the Sustainable Development Goals, but it promised to ensure the development of these technologies in an environmentally friendly way.

Following up on this promise, in 2021, a from the European Parliament strongly suggested to include environmental considerations in the list of potential risks of AI in — what at the time was — the Proposal for the Artificial Intelligence Act. This created a sense of hope that the would find methods to keep environmental harm caused by AI under control, but the EU might have fallen behind on its promise. 

… A Broken Promise?

One year after the publication of the official AI Act, the consensus among most experts and organizations is that the Regulation in question has failed to account for the many environmental risks of AI.

On one hand, the Act indeed sets standards that AI models must comply with to be put on the market. These standards include information on energy consumption, collaboration, and the possibility of setting up regulatory sandboxes for the purpose of developing more environmentally friendly AI systems.

On the other hand, however, there are multiple issues with the way the AI Act strives to tackle the sustainable side of AI. It has been argued that the sole focus on energy consumption fails to account for several different environmental concerns such as water usage, resource extraction, and e-waste. Moreover, the lack of strict mandatory requirements in favor of flexible standards creates a lenient framework of provisions that is hard to enforce.

This is what sparked the debate on the sustainability of AI regulation, with many arguing that the EU should do more to protect our planet from new technologies.

More Opportunities for Sustainable AI in the EU

The EU is starting to take additional steps to achieve a more environmentally-friendly AI regulatory framework. Indeed, the Commission announced its intention to publish a under the AI Act. A delegated act would allow supplementing or amending the Regulation, particularly by introducing the requirement of tracking the energy required to train the machines.

Even though this idea sounds promising, the focus is yet again on energy consumption. The EU must take into account the broader environmental impacts of AI to ensure a truly comprehensive approach to sustainable AI regulation.

To do so, the regulatory focus of the EU must address AI’s full environmental footprint by including setting strict standards for water usage, carbon emissions, resource extraction, and e-waste management. This should be paired with mandatory impact assessments and stronger enforcement mechanisms to ensure the effectiveness of the measures.

Despite being a frontrunner in AI regulation, the EU has neglected its sustainability promises in favor of the calls to AI investments. With those demands now met, it is important to revisit the regulatory framework to ensure these investments support the development of environmentally-friendly AI models.