AI Joins the Race with Climate Change
It is no news that climate change has been a topic of discussion for decades, and countries are attempting, or at least claiming to attempt, steps to mitigate and prevent it. There are also corporate actors furthering their own interests, focused on industrial expansion and innovation. While some states are running the climate race, some corporations and state-sponsored corporations are running the AI race and ordinary people, species, and the environment are caught between. Where innovation is important, there is also a flip side, there is always an environmental cost. The advocacy on reducing environmental pollution is well stated. However, converse to climate change, AI too has now joined the race, and this is the point. AI has grown to an unprecedented degree, so much so that its own creators question its survival. The world is no longer running one race against climate change. It is running two, on the same track, and the second runner is gaining.
AI has made life easy, and it is a fact. AI has become very convenient it makes day-to-day decision making easy, tasks delegated and timed accordingly, priorities checked and placed properly, and it helps in making informed decisions. But what is the real cost? The real cost, I mean not what we lose but, to be honest, the part that we do not know that we lost. When you use AI for day-to-day tasks, be it drafting a letter, planning a party, or venting emotions, AI gets a significant leverage about our life. It knows our ways, moods, choices, priorities and AI in fact guesses what is yet to come.
This is the new race. AI does not simply react. It anticipates. As a fact, AI is trained by humans, it mimics humans, and it develops through human engagement. However, when training an AI or feeding it information, humans once took the upper hand. Conversely, now AI has surpassed the human capacity to act first. And this is the paradox. AI acts ahead of us, yet it can only see backwards. It is proactive about a world it knows only through the data we fed it. Every prediction it makes is built from patterns of the past our past decisions, our past preferences, our past biases. So when AI shortlists a candidate, recommends a loan, flags a neighbourhood, or drafts a policy, it is not inventing the future. It is repeating our history, faster, and calling it foresight. And because it acts before we do, we stop noticing that the future it is building is assembled from the past that we handed over to it. This is the asymmetry. Not that AI is smarter than us — but that it moves first, using conclusions we never agreed to, drawn from data we never chose to give. This isn't something bad. However, this reflects dire urgency urgency where humans have to understand where to draw the line. Because AI is running the same race as climate change. It acts first, and it consumes first, energy, water, capital, attention, before we have decided whether it should. It draws on the same energy, the same water, the same capital, the same attention. It competes for the same political will that climate action needs. And while the first race is still unfinished, the second runner is learning our pace faster than we can set it, and this is not something to be celebrated.
The line does not mean stopping AI. Stopping AI is not the answer. The matter in question is to decide, fruitfully and sustainably, what decisions will be consulted and made through AI, what decisions need exclusively human decisioning, and what decisions need human intervention. Criminal verdicts, warfare targets involving civilians in war space, helplines, and such decisions cannot be autonomous only. The reason is that an algorithm may get the "right" answer faster, maybe the most logical answer, but legitimacy comes from the process, not the answer. It comes by weighing risks and benefits and consulting stakeholders. This is why the line must be drawn at delegation, not at use. We cannot stop AI from acting ahead. But we can decide which forward actions we permit it to take. If we let it act first on decisions built from our biased past, we have not delegated a task, we have delegated our judgement. Such intervention for certain things is required because AI works on the basis of pattern recognition and past decisions.
Take recruitment, if a company has always preferred men over women, and shortlisting is relied on AI, it will favour males only. So who will bear the consequence of this biasedness? The applicant who never learns why they were rejected, and who has no procedure to appeal. This is where the line becomes practical rather than poetic. It means audits, an independent check of what the system actually does, not what its makers say it does. It means a right to explanation, a rejected applicant is told why, and can contest it. It means liability that lands somewhere real, if an autonomous system cause’s harm, someone must answer for it, whether that is the developer, the deployer, or the institution that chose to rely on it. Also, legally speaking, "attribution" becomes tricky when it comes to autonomous decisions. Who made the decision? Is it the AI, and if so, is AI an agent, or is it the developer? Where is the command structure? This is therefore heavily contested. Irrespective of the conflict, the cost is borne by the person on the other side of the decision, who has no one to hold accountable. That is not a philosophical problem. It is a practical one, and it is already here.
Explaining further, data centres consume enormous electricity and water, which has a significant and undeniable negative impact. According to the United Nations (2026), "Data centres, the global infrastructure powering AI, could consume 945 terawatt-hours of electricity annually by 2030 — nearly triple the combined annual electricity use of Pakistan, Bangladesh, and Nigeria, countries collectively home to more than 650 million people… On top of the carbon footprint, every unit of electricity used by data centres also carries a 'water footprint' for cooling and energy production, and a 'land footprint' associated with power generation and supply chains." And the cost is not only in the training. The true price of convenience is paid not by the companies building the models, but by every user who asks a question, drafts a letter, or vents an emotion. Efficiency will not save us here either, because cheaper and faster AI drives more use, not less. This is the rebound effect, and it is the same paradox in a different form: the more capable the tool, the more we hand over to it, and the more it consumes. This showcases the accelerating repercussions of investing in AI without giving adequate attention to the environment. And this is where the argument has to name its price. If we want AI to be cleaner, someone pays — slower training, costlier models, higher prices passed to users, or fewer services offered for free. There is no version of this where the race is run at full speed and the environment is left unharmed. Choosing the line means accepting a cost, and saying openly who bears it. Where, of course, it is necessary to be technologically innovative and to predict and foresee threats using AI, it is essential to look into the other aspect and see what grievances are there.
The rationale behind these arguments is not short-sighted, because the effects of these unprecedented advancements have a direct bearing on the environment and its species in years to come. Right now, what people witness is convenient communication, creative production, threat analysis? Yet there is more beneath the surface. Water scarcity, soil degradation, loss of habitat, and the loss of dignity and privacy of individuals are also present. The truth is, it is not only economical costs, and there are also other human security concerns such as privacy, political freedom, social rights, and interests.
Conversely, there are also other matters. For example, it is important to reiterate that AI is not the enemy, or only a competitor in the race. It is also running for the other side. As per UN Environment Programme (2026), "Artificial Intelligence (AI) is helping turn satellite observations into methane mitigation action in the oil and gas sector. These mitigation efforts have delivered a climate benefit comparable to removing the annual emissions of almost 24 million gasoline-powered passenger cars." Further, satellites generate more data than humans can process. Therefore, AI processes more data than humans. It identifies confirmed detections before expert review. AI is used to predict monsoons for farmers, and AI is used to optimise power grids so less energy is wasted. Therefore, the reality is that AI is not inherently a climate villain. It is a tool. And if it is a tool, then the practical question is not whether it runs, but who is accountable for where it runs. That means naming the actor. National regulators, because they already license and restrict harmful technology. International bodies, because energy, water, and emissions do not stop at borders. And the companies themselves, because they build the systems and know their capabilities first. None of these alone is enough. All three, together, are the only version of this that is workable. This is what it means to set the finish line. Not a slogan, but a decision — made by someone, enforced somehow, at a cost we are willing to name. If we do not decide where the line is, the line will be decided for us ,by systems that only know what we have already done, never what we might still choose to become. The finish line is still ours to set. The question is whether we set it before the second runner crosses it.
Charani LCM Patabendige is a Researcher, Attorney-at-Law and a Visiting lecturer!
You can reach her via charani.patabendige@gmail.com