The impact of AI on the environment: a major issue... completely invisible
While artificial intelligence is invading media, economic and political space, its environmental impact remains virtually absent from the global conversation: since January 2024, less than 1% of messages published on X and LinkedIn have referred to it. A striking figure, revealing the marginalisation of ecology in digital debates. This invisibilisation cuts across all walks of life: the 100 world leaders in AI only rank climate as 8ᵉ of their priorities; communities committed to the ecological transition talk very little about AI (2.85% of exchanges); the media rarely address the environmental impact in their coverage of AI (only 2.4% of articles devoted to AI), with largely tech-friendly treatment.
This collective blindness - from innovators to ecologists, from editors to decision-makers - is holding back the awareness that is so essential to action.
Information fog and polarised narratives
In the absence of clear data, the debate is becoming tense, caught between a triumphant techno-solutionism and a radical rejection of technology, against a backdrop of climate disinformation: Grok, for example, slips climate sceptic arguments into 10% of its answers to climate-related questions, while 33% of French people still doubt the scientific consensus on global warming. The discourse is structured in closed bubbles, with no dialogue between promoters of all-powerful AI and those calling for digital degrowth. This polarisation is fuelling mistrust of science and hampering the development of collective, credible and ambitious solutions.
Measuring the invisible: urgency and challenges
We know that the digital footprint is growing twice as fast as that of other sectors (+5 to 6% per year), due in particular to the rise of generative AI. However, the methodologies for accurately quantifying the impact of AI remain fragmented: proprietary data, scope limited to GHG emissions, failure to take account of the rebound effect. However, open initiatives (Code Carbon, AI Energy Score, digital passport) and Open Data are providing a common basis for measurement and transparency.
Building a sober and ambitious third way
Faced with this "information fog", France is developing a pioneering ecosystem: from the AFNOR standard on frugal AI to measurement tools such as Code Carbon, via the Coalition for Sustainable AI launched at the Summit for Action on Artificial Intelligence. Concrete initiatives are emerging, from start-ups developing specialised models that consume less energy to companies integrating environmental criteria into their technological choices.
The conclusion of our analysis argues for a third way between techno-solutionism and radical rejection of technologies, favouring transparency, measurement and sobriety in the deployment of AI to serve the ecological transition.