During my conversation with Brockman and Sutskever, I had referenced one of them: the carbon footprint of training such models. In June 2019, Emma Strubell, a PhD candidate at the University of Massachusetts Amherst, had been the first to coauthor a paper showing that the footprint for developing large language models was growing at a startling rate. Where neural networks could once be trained on powerful laptops, their new scale meant their training was beginning to require data centers drawing significant amounts of energy from carbon-based sources. In the paper, Strubell estimated that training the version of the Transformer that Google used in its search for just a single cycle—in other words, feeding it some data and letting it compute a statistical model of that data—could consume roughly 1,500 kilowatt hours of energy.

Under Pinochet’s rule, Chile privatized nearly everything—education, health care, the pension system, even water. The strategy produced economic growth; it also fueled stunning inequality. Chile is among the most unequal countries in the world today, with nearly a quarter of the country’s income concentrated among a few powerful families in the 1 percent. Having never meaningfully industrialized, it also remains tethered to the extraction economy that makes it relevant to higher geopolitical powers.

And so, as the AI boom arrived, Chile would become ground zero for a new scale of extractivism, as the supplier of the industry’s insatiable appetite for raw resources, not just its copper and lithium in the north but also its land, water, and energy resources for a growing crop of data centers in the Santiago metropolitan region. In May 2024, the government proudly announced that the country would welcome twenty-eight new data centers, on top of its existing twenty-two, over the coming years, bringing in $2.6 billion of foreign investment.

That summer, as Google filed a report with Chile’s environmental agency for approval of its data center—a largely rubber stamp process—MOSACAT, a water activist group, began combing through all 347 pages of the filing. Buried in its depths, Google said that its data center planned to use an estimated 169 liters of fresh drinking water per second to cool its servers. In other words, the data center could use more than one thousand times the amount of water consumed by the entire population of Cerrillos, roughly eighty-eight thousand residents, over the course of a year. MOSACAT found this unacceptable. Not only would the facility be taking that water directly from Cerrillos’s public water source, it would do so at a time when the nation’s entire drinking water supply was under threat. In 2019, as with Iowa and Arizona, Chile was already nine years and counting into a devastating and historically unprecedented megadrought.

Pena had read about hyperscalers using potable water, even during major droughts, and the activism of communities like MOSACAT that had resisted the projects. But when he downloaded the details of the project, the water numbers were marked as confidential. After submitting a public information request, which he had successfully done around twenty times, the ministry continued to withhold the numbers, saying they were proprietary information. Pena wondered what they were hiding and worried about the precedent it would set for other cloud companies that would inevitably begin to eye Uruguay, following Google’s lead, for their own expansion. So he evoked the water clause in the constitution. With the help of a lawyer friend who was willing to work pro bono, he sued the ministry.

In March 2023, four months later, Pena won the case in a surprising victory. The environmental ministry revealed that Google’s data center planned to use two million gallons of water a day directly from the drinking water supply, equivalent to the daily water consumption of fifty-five thousand people. With much of Montevideo receiving salt water in their taps not long after, the revelations were explosive.

The bottleneck for producing novel biological weapons, for example, is not about finding a recipe, Weinstein noted. Such recipes already abound online and are easily found via Google. It is about obtaining the materials and equipment to actually make the armaments. Restricting access to so-called frontier models would thus do little to fix this. But the collateral damage of suppressing the publication of AI models would risk weakening the foundations of US AI innovation. Open source—sharing and building on code and software released to the broader community—has long been the bedrock upon which the wealth of US-based startups flourish. Restricting model weights from being published would give smaller developers fewer pathways than ever to create their own AI products and services. It would further entrench the dominance of the giants represented in the Frontier Model Forum. AI models would also become ever harder to scrutinize, such as in the work of Sasha Luccioni, Yacine Jernite, and Emma Strubell, who have relied heavily on open generative AI models to quantify the carbon and environmental costs of continuing to scale them.