This newsletter captures the increasingly widening scope of artificial intelligence – from the creative frontier of generative video to the moral dilemmas of autonomous warfare and the resetting of the carbon clock. As governments debate global regulation and companies dream of orbiting data centres, AI’s footprint is expanding far faster than the systems designed to keep it sustainable. What unites these stories is a shared tension: progress driven by innovation, yet shadowed by questions of environmental cost, governance, and control.

1. Sora 2 and the Hidden Cost of Creativity

OpenAI’s Sora 2 can now generate minutes of photorealistic video from a few lines of text – a leap that blurs the line between human and machine creativity. But the more realistic these outputs become, the more power they demand. AI already consumes about 20% of global data-centre electricity, a figure expected to double by the end of this year and potentially reach 100% of total data-centre use by 2028 (https://www.iea.org/reports/energy-and-ai/energy-demand-from-aI). Water demand is equally troubling, the global AI demand is projected to account for more than the total annual water withdrawal of Denmark or half that of the United Kingdom (https://arxiv.org/abs/2304.03271) while chip fabrication still requires massive volumes of ultra-pure water (https://www.weforum.org/stories/2024/07/the-water-challenge-for-semiconductor-manufacturing-and-big-tech-what-needs-to-be-done/)

Advocates argue the picture is overstated: hardware efficiency is improving each year and new data centres increasingly draw from renewable sources (https://www.linkedin.com/pulse/data-center-sustainability-strategies-2025-innovations-best-cetcc/) Yet every new generation of chips triggers another wave of resource use, from mineral extraction to cooling infrastructure – effectively resetting the carbon clock. As the 19th-century Jevons paradox showed, efficiency often leads to higher consumption, not less (https://energyhistory.yale.edu/w-stanley-jevons-the-coal-question-1865/).

For comparison, an hour of HD streaming on Netflix uses around 100 times more energy than generating a text-based AI response. But Sora 2’s move into long-form video could change that calculus quickly. Without transparency on energy and water use – ideally through mandatory, location-based reporting – it will be difficult to steer AI development toward regions rich in renewables and water resilience.

Sustainability aside, Sora 2 has also reignited debate about copyright and content ownership. The Guardian reports that studios accuse OpenAI of enabling widespread infringement as users replicate copyrighted characters. CEO Sam Altman has pledged to move to an opt-in model and offer creators greater control, but the episode highlights how quickly generative tools can test – and sometimes break – existing legal frameworks.

2. AI and the Sustainable Development Goals

AI’s relationship with the United Nations Sustainable Development Goals (SDGs) remains deeply conflicted. On one hand, the technology offers clear benefits: in agriculture, multilingual AI assistants now help millions of farmers each year by providing image-based advice on pests, soil, and weather conditions. The integration of voice and visual inputs has lowered barriers for those with limited literacy, while falling deployment costs have expanded digital access in rural areas.

Yet these gains exist alongside a widening digital divide. As computing power becomes the new global chokepoint, digital equity increasingly collides with energy justice. The infrastructure that enables AI’s promise is energy-intensive and unequally distributed, concentrating technological capacity and environmental burdens in a handful of nations.

To address these imbalances, the United Nations and its partners argue for integrating Indigenous knowledge systems, codifying ethical design principles, and recognising digital access as a civil right. Without these safeguards, AI risks deepening the very inequalities that the SDGs were created to reduce.

The reality is that AI will continue to evolve regardless of public hesitation. What matters now are the policy choices – where data centres are built, who controls the technology, and whose knowledge counts. This is what will determine whether AI’s expansion aligns with sustainable development or deepens existing divides.

3. AI on the Battlefield

The BBC reports that both Russian and Ukrainian forces are deploying AI across the front line – analysing more than 50,000 video streams each month to identify targets, map movements, and coordinate defences. In Ukraine, AI-guided drones can now lock onto a target and complete missions autonomously once launched, reducing the need for direct human control. Developers anticipate that the next generation of systems will be capable of independently locating and destroying targets.

While these technologies promise faster responses and bolstered air defence, they also raise grave ethical and humanitarian questions. Can algorithms reliably distinguish combatants from civilians, or friend from foe? What happens when machines are granted lethal authority without human oversight? The International Committee of the Red Cross (ICRC) warns that autonomous weapon systems challenge the foundational requirement of human judgement in the use of force and raise serious concerns about compliance with international humanitarian law. (ICRC) Legal scholars argue that such systems may undermine accountability, creating a “responsibility gap” in which neither programmers, commanders, nor machines can be held clearly liable. (OUP Academic) Ethical theorists caution that autonomous systems exacerbate risks to civilians, erode moral agency, and corrode human control over life-and-death decisions. (Taylor & Francis Online).Ukrainian engineers remain cautious, aware that over-reliance on automation could strip warfare of accountability altogether.

4. Space: The Next Frontier for AI Infrastructure

With energy demand from artificial intelligence growing exponentially, some in the industry are looking beyond Earth itself. Jeff Bezos predicts that within two decades, space-based solar data centres could outperform terrestrial facilities. In orbit, there is no night, no weather, and no need for cooling water — a near-perfect environment for the relentless power draw of large-scale AI models.

This vision reflects the scale of the challenge. According to the International Energy Agency (IEA), AI workloads could consume as much electricity as Japan’s national grid by 2030. Launching gigawatt-scale facilities into orbit would come with enormous costs — both financial and environmental — but it underscores how unsustainable current data-centre growth has become. As energy constraints tighten, the idea of space computing sounds less like science fiction and more like an inevitability.

5. The UN Steps In

Finally, the United Nations has launched a Global Dialogue on AI Governance alongside the creation of the UN Independent International Scientific Panel on AI (UNIISPAI) — the first body dedicated to assessing AI’s global social and environmental impacts. Announced during the 80th UN General Assembly, the initiative signals a shift toward collective oversight and shared responsibility.

The Dialogue will allow governments to address issues that no single state can manage alone – from algorithmic bias to energy use and emissions — aligning innovation with sustainability. A balanced framework depends on transparent audits, inclusive design, smart regulation, and international cooperation through the UNFCCC, G20, and IEA.

The UNFCCC’s “AI for Climate Action” report underscores both the promise and peril of AI – highlighting that efficiency alone isn’t enough. Sustainable AI must rest on precaution, inclusion, and systems thinking to ensure progress remains just and accountable.

Thanks for reading.