Ownership changes the balance
If a small number of organizations control highly capable AI, they may gain influence over production, information, security, and access to essential services. People who depend on those services can lose bargaining power even when the systems operate exactly as their owners intend.
This is a risk from concentrated human power. It should be examined separately from autonomous loss of control. A technology can remain obedient to its operator while helping that operator dominate everyone else.
Dependence becomes leverage
Suppose one provider becomes central to a country's businesses, public administration, and information services. Changing its terms affects many sectors at once. Alternatives exist on paper, but switching requires time, expertise, and infrastructure the customers lack.
Advanced surveillance, persuasion, or automation could deepen this imbalance if used to weaken opposition or remove opportunities to negotiate. The outcome depends on ownership, institutional safeguards, and the capabilities actually deployed. Powerful tools do not by themselves establish that an authoritarian outcome will occur.
Avoid solving one risk by creating another
Democratic oversight, limits on sensitive uses, competition, public accountability, and practical rights to leave a service can reduce concentrated power. Essential decisions need avenues for appeal and institutions capable of enforcing them.
There is also a real tension: distributing dangerous capabilities widely may reduce monopoly power while increasing misuse and proliferation. Keeping those capabilities tightly held may make monitoring easier while giving a few actors exceptional authority. A prevention policy must address both sides explicitly. The objective is to preserve broad human agency through accountable limits, rather than assume either unrestricted distribution or centralized ownership automatically makes people free.