"Generative AI's Hidden Costs: Energy, Trust, and Disinformation Risks"
Generative AI
AI systems that create new content (images, text, audio) by learning from large training datasets.
Diffusion model
A type of AI architecture that generates images by progressively refining random noise into structured output.
Generative AI refers to a class of artificial intelligence systems capable of producing new content, images, text, audio, and video, from learned patterns in large training datasets. Unlike earlier AI tools that classified or retrieved existing information, generative models synthesise entirely new outputs. The most widely used image-generation systems rely on diffusion models and transformer architectures, trained on billions of images scraped from the internet. This training process, and the inference (output generation) that follows, requires substantial computational infrastructure: large clusters of graphics processing units (GPUs), energy-intensive data centres, and significant water consumption for cooling.
The governance of AI has evolved rapidly at the international level. The Bletchley Declaration (2023), the Seoul AI Safety Summit (2024), and the AI Action Summit (Paris, 2025) have produced frameworks for responsible AI development. Yet these remain largely non-binding. At the Paris summit, two major powers did not sign the declaration on inclusive and sustainable AI, illustrating the limits of voluntary international governance. India's domestic AI governance rests on the Information Technology Act, 2000 and institutions such as CERT-In and NCIIPC, frameworks designed around human-directed threats rather than the subtler harms of autonomous or generative AI systems. The AI Impact Summit (New Delhi, 2026) produced the New Delhi Declaration, structured around seven pillars including Secure and Trusted AI, but binding domestic regulation of AI's social effects remains nascent.
The sustainability parallel the article draws is analytically important. For decades, the fashion industry externalised its true costs, water use, emissions, precarious labour, making cheap clothing appear almost frictionless to consumers. The sustainability movement forced a reckoning with what was hidden in the value chain. The same logic now applies to generative AI: a synthetic image that appears instantaneously on a phone is not consequence-free simply because it is intangible. The hidden costs include energy consumption, data consent, intellectual property (IP) rights of original creators whose work trained the model, and the gradual erosion of public trust in visual evidence.
Generative AI's social consequences operate across several distinct dimensions. The first is disinformation and epistemic harm. For most of the modern era, photographs carried a cultural authority rooted in the assumption that a camera had encountered something real in the world. Generative AI breaks that assumption. The consequence is not merely that people may believe false things, but that they begin doubting things that are true. This erosion of confidence in visual evidence is a systemic social cost, not an individual one. When AI-generated images enter political discourse, as fabricated street scenes or invented locations used as partisan attacks, the harm extends to democratic deliberation itself.
The second dimension is energy and environmental cost. Training a large generative AI model can consume as much electricity as hundreds of households use in a year. …
Challenges of digitalization
How can Artificial Intelligence (AI) and drones be effectively used along with GIS and Remote Sensing techniques in locational and areal planning?
Consider the following statements about the hidden costs of Generative AI:
1.The energy consumed in training and running generative AI models is a real environmental cost, even though the output (a synthetic image) is intangible.
2.Intellectual property questions arise because AI models are trained on creative works, often without explicit consent or compensation to original creators.
3.The IT (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 specifically mandate platforms to label or remove AI-generated synthetic media.
Which of the statements given above is/are correct?
Match the following AI governance milestones with their correct descriptions:
List I (Milestone) — List II (Description)
A.Bletchley Declaration — 1. Produced the New Delhi Declaration with seven pillars including Secure and Trusted AI
B.Seoul AI Safety Summit — 2. First major multilateral agreement on frontier AI safety risks
C.AI Action Summit, Paris — 3. A major power did not sign the declaration on inclusive and sustainable AI
D.AI Impact Summit, New Delhi — 4. Second in the sequence of international AI safety summits
Select the correct match:
Consider the following statements regarding the governance of Generative AI at the international level:
1.The Bletchley Declaration (2023) was the first major multilateral agreement specifically addressing frontier AI safety risks.
2.The AI Action Summit held in Paris produced a declaration on inclusive and sustainable AI that was signed by all participating major powers.
3.India's IT Act, 2000 was designed to address harms arising from autonomous AI systems generating synthetic media.
How many of the above statements are correct?
Which of the following best describes the concept of 'epistemic harm' in the context of Generative AI and synthetic media?
Want the full note, Prelims Facts & MCQ answers?
Create a free account — takes less than a minute.
Create free account