The same technology that monitors our forests consumes water, energy, and rare minerals on an industrial scale. With the start of COP30, Brazil must decide whether to regulate this paradox or pretend it does not exist.
Artificial intelligence systems like PrevisIA from Imazon and the Inpe platform detect deforestation in real time, predict where degradation will occur, and map secret roads where inspectors have never set foot. This digital sentinel will protect the Amazon with efficiencies that states could never achieve on their own. However, according to Stanford University’s AI Index Report 2025, the Rama 3.1 405B training produced 8,930 tons of CO₂. This equates to 900 flights between São Paulo and New York.
The real impact is in the reasoning. Millions of users browse ChatGPT every day, generating emissions that can exceed the overall cost of initial training. Every 20 to 50 questions consumes 0.5 liters of clean water. Data centers rely on lithium and tin extracted from degraded ecosystems.
At this COP, Brazil needs to present concrete legal solutions rather than speeches. And here is our contradiction. While we monitor the Amazon with sophisticated AI, we operate 180 data centers across the country, including Barueri (SP), the fifth largest in Latin America, without a regulatory framework that requires specific environmental controls for this infrastructure. The metaphor of “cloud computing” has created an illusion of immateriality. Reality is physical. Each model requires thousands of refrigeration servers that drain the aquifer. These servers are often powered by energy derived from fossil fuels and are built with minerals extracted without a focus on sustainability.
Our strong environmental laws were designed in the analog era. We couldn’t imagine regulatory algorithms that encourage unsustainable consumption, data centers that consume as much water as a small city, or AI models that can be trained to produce emissions equivalent to hundreds of transatlantic flights.
We can demonstrate our ability to innovate legally by hosting the most important climate conference on the planet in the heart of the Amazon. Brazil brings together unique elements, including having the largest biodiversity heritage site on earth, operating advanced AI systems to monitor it, and hosting critical data center infrastructure. This combination allows us to propose a governance model that other countries can emulate. However, this requires urgent legislative action.
First, we classify data centers and AI infrastructure as potentially polluting activities and subject them to specific environmental licenses. Second, establish radical transparency. Developers must publish detailed reports on the environmental footprint of their models.
COP30 could become a forum for Brazil to propose international standards for these indicators.
Third, it incorporates the guardian-recipient principle and creates incentives for companies to develop AI with positive environmental externalities. Fourth, we will develop an “Environmental AI Charter” that integrates duties, responsibilities, and incentives.
Regulatory inertia is not neutral. It allows technological progress to continue without internalizing environmental externalities and transfers its costs to the next generation. AI could be Amazon’s most powerful ally, but only if there is legal wisdom to transform it into intelligence for sustainability.
COP30 is more than just a conference. This is the deadline we have set for ourselves to decide whether Brazil will be a leader in environmental governance in the 21st century or another country passively watching its own contradictions.
Trends/Discussions
Articles published with signatures do not reflect the opinion of the newspaper company. The aim of this publication is to stimulate debate on Brazilian and world issues and to reflect different trends in contemporary thought.
About The Author
