NPC! The New Currency of the Ultra-Rich

1 Who appointed the ultra-rich to decide for humanity?

When someone with immense wealth takes it upon themselves to decide what humanity needs, they are claiming a mandate. The people expected to live with those decisions have every right to ask who gave it to them.

Ownership can give someone control over working conditions, access to a market or the terms of a service. People can agree to those terms because an arrangement serves their own purposes. Expertise and delegated decision-making can make it work. The concern arises when their cooperation or exposure to risk becomes useful to others while their ability to shape the arrangement is weak.

Formal permission to leave supplies only part of that assessment. A person may be free to walk away on paper and still lack an affordable alternative, continuity of an essential service or the resources to sustain a challenge. Accessible alternatives and a timely, independent remedy weaken the case that people are subject to someone else’s terms.

In 2017, Facebook founder Mark Zuckerberg assigned his company a public role. It would build what he called “social infrastructure” for a global community, including tools for civic participation and new ways to make collective decisions. He also acknowledged that executives in California were poorly placed to set standards across cultures. [3]

After the attack on the US Capitol on January 6, 2021, Facebook blocked then-President Donald Trump from posting on Facebook and Instagram and extended the block indefinitely. Its Oversight Board upheld the restriction on safety grounds and required Facebook to reassess the indefinite penalty under its own rules. The company could shape public debate through its control of the platforms. The board could constrain how it exercised that power. [3]

Lobbying offers another route into public decisions. Richard Hall and Alan Deardorff describe lobbying as legislative subsidy: organized interests can supply sympathetic lawmakers with information, policy expertise and legislative labor. Influence can operate through the capacity to develop and advance an agenda, even when nobody’s convictions change and no vote is purchased. [4]

A citizen and a corporation may therefore both have a right to express an opinion while possessing very different capacities to keep that opinion present in public decisions. The question is whether people who will absorb the consequences have a meaningful opportunity to shape, oppose or revise the proposal.

The Washington Post’s top-50 tally put Musk at $446 million and Republican-leaning and Democratic-leaning donors at roughly $1.7 billion and $830 million, respectively. [5] Separately, the Brennan Center reported $483 million for Musk and United States of America Inc. combined. [6]

Many small contributions can also give campaigns resources. The Washington Post found Democratic congressional campaigns received $574 million from donors giving under $200 in total, compared with $143 million for Republicans. [7] Those figures track contributions to campaigns. Spending by outside groups requires separate accounting.

James Talarico defeated Ken Paxton in Texas despite the substantial outside effort backing Paxton. In Nebraska, wealthy incumbent Pete Ricketts defeated Dan Osborn, a mechanic and former union leader. Both Nebraska campaigns raised substantial sums, and Talarico had significant resources of his own. The cases show different outcomes under substantial spending. Isolating its influence would require evidence that separates campaign resources from incumbency, party allegiance, candidate appeal and organization. [8, 9]

The inquiry then follows money into particular decisions. Did it help someone keep an issue on the agenda, obtain access to decision-makers or set terms others had little opportunity to challenge? An election result alone leaves those questions open.

2 Democracy changes the game. NPCs become players.

Democracy gives people ways to challenge the terms they live under and replace the people who set them. When they organize, vote or stand for office, the people cast as NPCs can change how the game is played.

Power exercised through other people’s participation depends on that participation continuing. Institutions need workers, customers, students, administrators and voters. Their cooperation can be constrained, encouraged or rewarded, but it cannot safely be treated as a permanently predictable feature of the environment.

Collective organization can change the available choices. A grievance that an isolated person cannot sustain may become a demand supported by institutions, legal assistance, shared resources or an electoral alternative. The people concerned exercise a capacity they already possess; recognition by those in authority follows rather than creates it.

Serbia’s student-led movement developed after the November 2024 Novi Sad railway-station canopy collapse, which killed sixteen people. Organization continued beyond the initial protests and became an electoral challenge. The final results of the October 25, 2026, election gave the Students’ List 140 of the 250 seats in parliament. The electoral commission allocated the mandates on November 6. [10]

The significance here is the work of turning participation into an alternative capable of contesting authority. The result gives no guarantee about subsequent government. A movement that acquires public power assumes responsibilities toward people who supported it and people who did not.

In the United States, November’s results gave Democrats majorities in both chambers, with some races still uncalled on November 11. The president’s party usually loses seats at midterms. That baseline helps place the result in context. A fuller explanation would have to distinguish voter priorities, candidate appeal, fundraising and organization. [11]

Agency also includes choices I oppose. Voters who return a wealthy incumbent may be acting on priorities of their own. Voters who choose a challenger authorize that person to hold office within its limits. I have to apply the same standard of agency to both.

Organization need not culminate in winning an election. Elinor Ostrom’s work on collective governance documents people developing rules for managing shared resources. Such arrangements challenge an assumption of permanent passivity by showing how participants can establish responsibilities and constraints among themselves. [12]

Collective purpose also creates opportunities for entitlement. A movement may have a defensible grievance while a faction assumes the right to decide which risks all participants, residents or workers must bear. A claim to liberation still needs authorization for the actions taken in its name.

Moral sincerity does not settle that authorization. People can sincerely believe that their cause warrants others’ sacrifices. Responsibility must follow actual authority, knowledge, discretion and foreseeable consequences, including the ability of participants to disagree, replace representatives or refuse particular actions.

Underestimation is a proposition to investigate. Establishing it in a particular case requires evidence about expectations and their effects on decisions. The narrower point is already consequential: power that relies on other people’s continued cooperation contains a relationship those people may be able to change.

3 Starlink, Starshield and the life-or-death power of unelected decision-makers

A decision to restrict communications that a country relies on for its defence can cost lives. A government can legitimately delegate operational decisions to a private provider. The democratic test is whether someone answerable to the public can review a restriction and reverse it in time. Even where a contract gives a government grounds to object, the people relying on the connection bear the immediate risk.

In October 2026, Musk acknowledged a speed restriction on civilian Starlink in Ukraine, citing terms governing offensive use. Ukrainian officials disputed the alleged breach. Responsibility for restrictions on Starshield, its military counterpart, remains disputed. Reuters cited officials who traced an order to the White House. The Wall Street Journal reported that Musk authorized the slowdown. Axios relayed The Telegraph’s account of Musk’s proposal, cited a source who said Trump raised no objection, and reported a White House denial. [13]

Proposal, authorization and implementation can involve different actors. The available accounts leave the full decision chain unresolved. The distinction matters because responsibility follows what each actor had the authority and discretion to decide.

On October 24, Ukraine reported that both services had returned to their previous speeds. Its defence minister credited discussions with Washington. This was a route to correction that worked through government intervention, after disruption had already affected users. The reporting leaves open what substitutes users could have deployed in time and what contractual means they had to challenge the change themselves. [13]

The case shows how dependence can separate the people bearing immediate costs from those able to alter the terms. A judgment about entitlement also needs the relevant decision rights, practical alternatives and means of review. The same questions apply when the service delivers real benefits and when its provider has legitimate terms to enforce.

4 When AI denies you a service, who can overturn the decision?

The institution using AI to decide who qualifies for a service owes applicants an explanation for a denial and a route to someone with the authority to reverse it.

AI makes the authority to assign objectives particularly consequential. A technically reliable system can pursue an objectionable human purpose accurately. A defensible purpose can also be undermined by a system that misunderstands instructions, optimizes an inadequate proxy or behaves unreliably. These are distinct failures requiring different evidence.

Consider an agency using AI to allocate a scarce service. A system might apply an exclusionary eligibility rule exactly as instructed, placing the problem in the human policy. It might instead misapply a defensible rule and deny eligible people. A further failure would arise if applicants could not discover or challenge the basis of a denial. This hypothetical separates the authority behind a decision, the system’s behavior and the standing of those affected.

Keeping a human in control addresses only part of the arrangement. The human may possess a valid mandate, exceed it or pursue an institutional objective that leaves affected people little influence. Governance must examine who authorized the objective, the evidence of reliable operation and the means to correct consequential decisions. NIST’s voluntary AI risk framework treats governance, design, deployment and evaluation as connected responsibilities. [14]

The original NPC problem can therefore persist inside a well-functioning technical system. Workers, applicants or users may be represented principally as variables through which somebody else’s objective is optimized. Greater efficiency can enlarge the capacity to act without enlarging those people’s ability to decide what is being done with their participation or information.

Human objectives deserve examination in their own right. They cannot be treated as a single public interest simply because humans supplied them. The companies, governments and departments giving instructions have different purposes and different claims to authority. Equally, institutional accountability cannot substitute for evidence that the technology operates safely and reliably.

People have purposes of their own, including when their work, information or vulnerability is useful to a larger undertaking. Their agency matters when it makes someone else’s plans harder to carry out. Those who challenge an assigned role take on the same responsibility toward others if they gain the power to assign roles themselves.