Investigation | Power & Accountability
By Andrew McDonald · 10 July 2026 · Analysis
Artificial intelligence could improve healthcare, productivity and access to knowledge. But its development is exposing a harder question: who controls the technology, who receives its benefits, and who carries its costs?
A worker is told that AI will make her more productive, but there is no guarantee she will share in the gains. A woman discovers that her face has been placed into synthetic sexual material. An Australian household is promised cheaper, cleaner electricity while billion-dollar data centres prepare to draw more power from the same grid.
These are different harms, but they reveal the same imbalance. The companies and institutions deploying AI can move quickly. Workers, regulators and communities are left to negotiate the consequences after deployment has begun.
Control Is Concentrated
The AI industry is not an open contest between equal participants.
Stanford University’s 2026 AI Index reports that industry produced more than 90 per cent of notable frontier AI models in 2025. It also found that the United States hosts 5,427 data centres, more than ten times the number in any other country, while one Taiwanese company, TSMC, fabricates almost every leading AI chip.
This does not mean a single company or country controls AI. It means that critical parts of the system, including advanced chips, cloud services, data centres and model development, are concentrated among a limited number of powerful organisations. The OECD has separately warned that cloud computing markets have high concentration, barriers to entry and difficulties for customers seeking to switch providers.
That concentration matters because organisations controlling the infrastructure can influence the price, availability and rules of access to increasingly important technology.
Workers Face a Transition, Not a Guaranteed Windfall
AI will not simply eliminate every exposed job. In many occupations, it is more likely to change tasks, increase output or assist workers.
The International Labour Organization estimates that AI may affect nearly 80 million workers across Southeast Asia, although it says large-scale job disruption has not yet occurred.
In Latin America and the Caribbean, World Bank and ILO modelling found that generative AI could improve productivity in 8 to 12 per cent of jobs. However, up to 17 million of those jobs may be unable to realise the benefit because of inadequate digital infrastructure. The study also estimates that 2 to 5 per cent of regional jobs face potential automation, with women twice as likely as men to be in that category. These are estimates of exposure and potential impact, not confirmed job losses.
The missing element is a fair bargain. Large and well-funded employers can purchase automation, while workers are often expected to fund their own retraining and accept greater uncertainty. Higher productivity does not automatically produce higher wages, reduced working hours or improved job security.
Synthetic Content Is Creating Real Victims
AI-generated content is no longer merely a problem of fake celebrity photographs or misleading advertisements.
UN Women reports that legal systems and platforms are failing many women subjected to AI-enabled deepfake abuse. Europol has also coordinated an international operation involving authorities from 19 countries that resulted in 25 arrests connected to AI-generated child sexual abuse material.
From 2 August 2026, European Union transparency rules will require clear labelling in key cases involving deepfakes, interactive AI systems and AI-generated or manipulated text concerning matters of public interest.
Labelling is necessary, but it is not a complete remedy. A label cannot reliably reverse reputational damage after fabricated material has spread.
The pressure is also reaching science. On 7 July, Nature reported on an academic “humanizer” designed to remove apparent signs of AI use from research papers and grant proposals. AI can help researchers analyse information and communicate findings, but tools designed to conceal its use undermine disclosure and make already strained review systems harder to trust.
Australia Will Feel the Physical Cost
The cloud is physical infrastructure.
The International Energy Agency reports that global data-centre electricity use increased 17 per cent in 2025, while use by AI-focused centres grew 50 per cent. Its central projection has total data-centre consumption rising from 485 terawatt-hours in 2025 to 950 terawatt-hours in 2030.
In Australia’s National Electricity Market, average data-centre demand was almost 600 megawatts during the first quarter of 2026. AEMO said 11 proposed centres, representing a possible ultimate load of 5.4 gigawatts, were progressing through transmission connection processes. AEMO cautioned that projects are staged and some seek multiple connection options, but expects its demand forecasts to rise materially.
AI infrastructure can support investment and productivity. The public should still know who will pay for the required generation, networks and storage, and how reliability and household costs will be protected.
The Standard Must Be Human Benefit
AI policy is both a technical and political challenge. Technical safeguards matter, but decisions about wages, infrastructure costs, competition and legal rights are choices made by institutions.
Workers need credible transition plans. Victims of synthetic abuse need fast legal remedies. Communities need transparent information about data-centre power and water use. Governments need independent expertise and the ability to enforce rules against the companies they regulate.
The question is not whether AI will become more capable.
It is whether people will retain enough power to determine what those capabilities are used for.
Without that, AI may improve productivity while automating inequality at scale.
Continue the investigation: The Data Centre Community Impact Hub brings together the evidence, US project tracker and practical tools for communities assessing data-centre proposals.
Related investigation: Meta Needed a Data Centre. BlackRock Found the Money. Who Carries the Risk? examines how the physical and financial burden of AI infrastructure is distributed.
Editorial note: This article synthesises published reporting, official statements and research. It is analysis, not original field reporting.
Editorial disclosure: This article was developed with assistance from artificial intelligence. Its sources, claims and conclusions were reviewed and approved by Immortal AI’s editor.
Leave a Reply