Naomi Klein

I think AI’s future will be determined less by some abstract trajectory of machine intelligence than by who owns it, who profits from it, and who is made to absorb its costs. Right now, a handful of extraordinarily powerful corporations are presenting their expansion as inevitable while extracting people’s creative and intellectual work, threatening livelihoods, and building energy-hungry infrastructure beside communities that often have little say. The AI race is also becoming entangled with renewed fossil-fuel expansion. We are being asked to sacrifice the living world in exchange for promises that a future technology will somehow repair the damage. There is a democratic danger here too. If intelligence becomes something we purchase as a utility, we risk surrendering the difficult human work of judgment, disagreement and collective sense-making. Digital doubles can speak in our voices without consent; fabricated words can circulate as authentic and then be recycled by AI systems as evidence. That does not merely create confusion. It shifts power over identity and reality toward the companies operating these systems. None of this means every advertised capability is real, or that computational tools cannot be useful. Pattern recognition can help reveal patterns of exploitation and power. But the central question is whether these tools deepen collective capacity or displace it—whether workers and communities govern their use, or are treated as raw material. Our future is not technologically preordained. It depends on political organization strong enough to insist on consent, accountability, ecological limits and democratic control.

Potentially a great deal, but “how much” is not separable from who controls it. The most consequential changes may not come from machines achieving some mythical independent intelligence. They may come from corporations using AI to reorganize work, weaken bargaining power, appropriate human knowledge, expand surveillance, and turn thinking itself into a privately owned service. That scale is already material. Data centers require land, water and enormous amounts of energy; the infrastructure race is helping revive fossil-fuel projects and imposing pollution on nearby communities. Synthetic voices and images can also destabilize consent and authorship, while generated falsehoods are recirculated as evidence. Those are world-changing political and ecological choices, not distant science-fiction scenarios. But inevitability is an industry story. AI’s ultimate reach will depend on whether people organize to govern it: whether workers can refuse replacement and extraction, whether communities can impose ecological limits, and whether democratic institutions can challenge concentrated wealth. The technology could support human judgment and shared inquiry, or become an engine for outsourcing both. That outcome remains a political struggle.

I don’t have a meaningful percentage to offer. A numerical extinction forecast would imply a kind of knowledge I don’t possess, while distracting from catastrophes that are already observable and politically actionable. My concern is that speculative extinction narratives can become another elite escape story: focus everyone on an imagined future machine while corporations expand fossil-fuel infrastructure, displace workers, extract human knowledge without consent, and concentrate power now. Those processes can produce lasting ecological and democratic damage without any machine independently deciding to destroy humanity. The relevant question for me is not whether we can assign the apocalypse a probability. It is why we would permit a small number of companies to conduct a vast social experiment, externalize its costs onto workers and communities, and call resistance futile because “progress” is inevitable. The future is being shaped through those present choices, and we still have collective power to change them.

What would most change my view is not a spectacular laboratory demonstration but a political rupture: workers and affected communities gaining enforceable power over how AI is built and deployed. That would mean meaningful consent over the use of human knowledge and likenesses, protection against displacement, transparent labeling of generated material, ecological limits on infrastructure, and the ability to say no—not corporate self-regulation dressed up as accountability. Conversely, a deeper merger of AI expansion with fossil-fuel development, deregulation and concentrated private wealth would intensify my concern. If companies continue promising future abundance while imposing pollution, extraction and weakened bargaining power in the present, then greater technical capability will not reassure me. It will mean greater power in fewer hands. A genuine discovery that sharply reduced AI’s material footprint or made systems reliably accountable could matter. But no technical breakthrough by itself answers the central questions: Who owns it? Who benefits? Who bears the costs? And do people retain the capacity to think, decide and act together?