A simple search in ChatGPT unleashes the immense power, unbeknownst to most users around the world, to absorb and devastate both natural and human resources. The numbers are staggering: a single data center can consume as much water as 360,000 homes, while up to two million books may have been destroyed by the AI company Anthropic to train its language models. As the UN University report published on June 3 argues, AI systems pose “existential risks”, from the depletion of resources to their influence and use in warfare.
The warnings from this UN institution coincide with the first encyclical by Pope XIV, Magnifica Humanitas, which concerns itself, among other things, with the environmental impacts on our “common home”. The two landmark documents implicitly engage with the concept of the commons, which refers to the collective actions to care for shared resources in an equitable manner, to sustain them. The destruction and appropriation of the commons for and by AI can be understood as biokleptocracy: a regime based on seizing life-sustaining natural and human resources for self-serving technological advancement.
The notion of the commons implicates a struggle against a hypercapitalism that views resources such as land and water as private goods to be extracted and monetized, excluding affected communities from decision-making; and hypermodernism, which drives overconsumption by exhausting resources to feed impetuous technological acceleration. According to the same UN University report, last year, all data centers underpinning artificial intelligence consumed as much electricity as Argentina, Chile, and Colombia, combined. Behind the impressive feat of a large language model such as ChatGPT polishing an email for us, a data center campus spanning about 10 city blocks can emit waste heat equivalent to that of 200,000 households, causing temperatures to rise in its vicinity.
So far, efforts to measure AI’s environmental impact have focused on emissions. The new UN report stresses that this metric misses the full environmental costs. Where data centers are fueled by renewable energy instead of fossil energy, they can actually put further pressure on local resources. Brazil’s solar and wind resources have attracted attention from data center companies looking to reduce emissions, and the country’s AI plan includes investments in renewable power for data centers. Yet already, Brazilian renewable energy projects have caused local deforestation and the loss of farmland.
The allure of bringing investment and creating employment, coupled with the promise that the same AI can contribute to optimizing energy consumption, may compel local governments to authorize data center campuses without considering the consequences. Industry standards recommend temperatures up to 27 °C for optimal cooling conditions, but there are data centers in 21 countries in places where this limit is exceeded, including in Brazil and Chile, which have experienced a boom in data center construction. AI systems functioning under heat stress conditions require additional electricity and water for cooling, further exacerbating Brazil’s water scarcity and Chile’s mega-drought.
Brazil and Chile also have some of the largest reserves of critical minerals in the world, including gallium, a byproduct of massive copper and aluminum mining, which is in high demand to replace silicon in semiconductors. Not only do local communities carry the burdens of hosting data centers, but they also face impacts from mineral extraction to build the technology, and from the disposal of electronic waste once the technology’s lifespan is finished.
Taking Garrett Hardin’s “tragedy of the commons” further, AI companies plunder human knowledge and ideas to sell them back to us, devaluing and homogenizing them in the process. AI converts human thinking into a mimicry that appears diverse only on the surface. A 2024 study found that individually, short stories written with AI assistance were evaluated as more creative. Yet taken as a whole, the AI-supported stories were similar, as opposed to the diversity of human-written stories. Such generative AI platforms are fed with the human ideas they then flatten: Anthropic cut the pages from books to scan them into a private collection, trashed the originals, and used the digitized copies without consent from the authors.
This biokleptocracy driven by AI does not stop with the natural resources used to build and run its hardware, or the human resources used to train its software. AI is pushing the renewed race of planetary and spatial conquests, this time led by commercial actors with limited oversight and restraints. Leaders of universities, in theory custodians of knowledge, show inaction toward or even affection for AI, leaving the educated population unequipped to recognize or address these spiraling existential threats.
For someone enthusiastic about the promise to extend life through parabiosis, or transfusing younger people’s blood into his own veins, Peter Thiel is ensuring no place is left to live in plenitude on this planet. Among other ventures, Thiel’s Palantir AI software enabled Israel’s genocide in Gaza, having taken the name of the crystal ball in The Lord of the Rings for the company.
From appropriating land to fueling genocide, AI creates a biokleptocratic cycle in which the destruction of common resources and ideas only escalates, without multilateral institutions actively engaged in creating a strong governance framework.
Autor
PhD in Health Promotion. Member of the International Advisory Board of The Lancet Global Health and member of the Steering Committee of the Thematic Working Group on Health Systems in Fragile and Conflict-Affected Settings of Health Systems Global Health.
Environmental journalist and master in Fine Arts (MFA) in fiction at the University of New Orleans and training at the Research Journalism Center (SUJO) in Bergen, Norway.
