A coalition of non-governmental organizations accuses companies like Google and Microsoft of green self-promotion to hide the high environmental costs of generative AI.
The platforms analyzed 154 statements “claiming that AI will have a net climate benefit, including those from companies like Google and Microsoft and institutions like “.
Lack of scientific evidence
According to the NGOs, this is the first time that the argument that AI could offset the increase in demand for fossil fuels generated by data centers has been critically analyzed.
“Only 26% of claims cite published academic articles and 36% cite no evidence at all. Overall, these claims are based on weak evidence, not peer-reviewed studies,” they emphasize.
Criticism of Google, Microsoft and the International Energy Agency
The report criticizes the Google for stating in official documents that AI could mitigate between 5% and 10% of global emissionsan estimate based on extrapolated data, without scientific basis, from a private consultant.
The study also adds that the multinational attributes to AI benefits that, in fact, are the result of solar panels and not artificial intelligence.
A International Energy Agency (IEA), the document points out, assumes that “the benefits far outweigh the direct emissions”, based on theoretical models rather than empirical measurements, and also uses vague and poorly substantiated claims that AI will enable savings of “up to 50% of waste”.
Furthermore, the AIE indicates, in some of its own studies on the subject, that they were reviewed by experts linked to technology such as Google, Amazon, Nvidia, Meta and Microsoft, something that NGOs consider a potential conflict of interest.
In turn, the Microsoft says it is working on generative AI to “empower a sustainable workforce”, without verifiable data or quantification of emissions reductions, the report said.
The analysis nto the found “a single example in which generative systems (…) such as ChatGPT, Gemini or Copilot have generated material, verifiable and substantial emissions reductions”, point out the authors.
Generative AI different from “traditional” AI
The study concludes that The statements serve to hide the differences between generative AI, which has high environmental costs, and “traditional” AI, which has a much smaller environmental footprint, for example, in forecasting wind patterns.
“This misleading change is a new form of ecological self-promotion,” the NGOs claim.
AI Impact Summit
The study, led by climate and energy analyst Ketan Joshi, was published on the eve of which takes place on Thursday and Friday in New Delhi.
“It appears that technology companies are using the lack of clarity about what happens inside data centers that consume enormous amounts of energy to hide a destructive expansion for the planet,” said Joshi.
The expert added that “promises of life-saving technologies remain hollow while data centers continue to fuel coal and gas every day.”
