AI oligarchy: In Love, Death & Robots, a sentient yogurt takes executive power over the United States after proving more competent than its human rulers. Many viewers saw a joke about rogue AI. Jean-Paul Carvalho, an Oxford political economist, offers a sharper reading. For him, the yogurt represents the elite created by AI; the machine is only the tool.
That shift in focus is useful. Much of the AI debate remains stuck on alignment, sentience and killer machines. A more plausible political risk lies nearer home. AI and humanoid robotics may weaken the institutions that gave ordinary citizens bargaining power in the 20th century, including public education, labour rights, welfare states and representative democracy.
READ | AI divide threatens India’s higher education goals
AI and the old social contract
Past technologies remade society, but they did not abolish the need for people. Settled agriculture produced large states, writing, taxation and organised religion. The Industrial Revolution produced factories and trade unions. Even the feudal lord needed peasants, clerks, smiths and soldiers. Industrial capitalists needed workers and managers. Power was unequal, but rulers and owners could not dispense with human labour.
AI combined with humanoid robots may alter that bargain. Mustafa Suleyman, co-founder of DeepMind and CEO of Microsoft AI, has proposed a benchmark for artificial capable intelligence. Such a system would take a $100,000 investment, study a market, design a product, contact manufacturers, negotiate prices, manage logistics and collect revenue without human direction. Suleyman has said this could happen by 2028.
In that production chain, the owner of the $100,000 remains essential. The analyst, designer, negotiator and logistics manager become software. Financial capital gains power as labour loses indispensability.
Ordinary employment shocks can be managed through retraining, welfare and new industries. The larger issue here is the erosion of mutual dependence. Public schools, public hospitals, welfare states and universal suffrage arose from hard bargains. Labour and human capital were indispensable to production, so elites had reason to pay taxes and accept mass politics. If machines provide both cognitive and physical labour at scale, that reason weakens.
READ | India’s AI challenge is a workforce skilling challenge
AI taxation and state power
State finance then becomes vulnerable. If tens of millions of wages disappear, income tax receipts shrink. Governments would need to tax the AI sector itself. The firms that control the technology would then hold leverage over public revenue and production.
Carvalho’s stronger warning is about dependency. A country reliant on a handful of AI providers can be hurt without tanks or embargoes. Denying access to models could paralyse firms and public agencies built around them. Carvalho gives the example of a 1% degradation in model quality compounded over a decade. The loss would show up in productivity, growth and competitiveness, without an obvious culprit.
Sovereign AI and developing economies
A world in which frontier AI is concentrated in Silicon Valley and Beijing leaves middle powers such as Britain, Germany, India and Brazil with a hard choice. They can buy access or spend heavily on sovereign AI systems.
For poorer economies, the problem is worse. Affordable labour has been the ladder for industrialisation. If humanoid robots and AI systems replace that advantage, many developing countries lose the one asset that allowed them to compete. The leverage of AI powers could exceed that of oil producers in the 20th century. Oil producers controlled energy. AI powers could control cognitive labour.
READ | Public AI infrastructure can give India its next UPI moment
Antitrust and AI concentration
Carvalho presents this as worst-case thinking. It is still useful because choices on competition law, public investment and democratic safeguards are being made before the technology settles into a stable ownership pattern.
Antitrust has to move beyond the narrow consumer-price test that dominated from the 1980s. The older American trust-busting tradition asked whether private concentration could coexist with democratic government. Senator John Sherman warned that a political system rejecting kings should not accept a king over the necessities of life. The same principle applies to AI systems that may supply the world’s cognitive labour.
AI oligarchy, need for governance and democratic power
Democracy also needs institutional reinforcement. Germany’s co-determination system puts workers inside boardrooms when automation decisions are made, even if it cannot solve ownership concentration. Tax policy should follow productivity gains where they arise. If AI replaces payroll income with rents, governments will have to tax those rents.
Carvalho also revives sortition, the random selection of legislators from the general population. Ancient Athens and medieval Venice used it to limit capture by organised factions. Sortition is no cure. It shows how older democracies tried to stop institutions from being bought.
None of this requires hostility to AI. The technology could raise living standards, speed up research and reduce drudgery. The issue is ownership and political power. A society that lets a few firms control cognitive labour will not remain politically equal for long.
We are still partly behind the veil of ignorance. We do not know who will own the main systems, who will gain from them and who will be made redundant. That uncertainty is useful. Rules written before fortunes harden are easier to enforce.
The yogurt is already here in outline. If the rest of society waits until it is redundant, it will bargain from weakness.
Susanthika S is Assistant Professor, Christ University Delhi NCR Campus. Dr Loganathan S is an independent researcher based in Noida.