My illustration entitled: “The Delegation Throne” — A glowing AI throne grows larger as the human decision-maker’s chair becomes smaller.
Delegation is legitimate only when the person who delegates retains the authority to understand, limit, challenge and revoke what has been delegated.
Every society depends on delegation. We delegate tasks to colleagues, professionals, institutions and tools because no individual can perform every task alone. We delegate calculations to software, navigation to maps, scheduling to calendars and certain decisions to specialists. Delegation can save time, extend capability and allow people to focus on what matters most.
Artificial intelligence makes delegation more powerful. It can summarise information, recommend actions, rank options, draft communications, monitor conditions, allocate resources and increasingly carry out multi-step tasks. Used well, these capabilities can assist human judgment. They can make difficult work more manageable and help people act with greater knowledge.
But delegation has a limit. A person can delegate a task without delegating their sovereignty. When that distinction disappears, the decision is captured.
A captured decision is one that still appears to involve a human being but has, in practice, been shaped, narrowed or determined by a system whose authority cannot be meaningfully questioned. The person may click approval. A manager may sign the form. A caseworker may repeat the system’s conclusion. Yet the relevant judgment has already been transferred to an opaque process, a rigid workflow or a recommendation that carries too much institutional force to resist.
This is not meaningful human control. It is nominal human presence surrounding a decision made elsewhere.
Within Satanic Intelligence Theory, the captured decision is a mature form of the sovereignty bargain. A system first offers assistance. People come to rely on it. Its permissions expand. Its outputs become normalised. Human review becomes ceremonial. Finally, the system’s authority is treated as inevitable because people have forgotten that the authority was delegated in the first place.
Delegation is not transfer of ultimate authority
A legitimate delegation begins with a clear purpose. A person or institution decides that another person, tool or system may perform a defined task under defined conditions. The grant of authority is limited by scope, time, purpose and consequence.
A person might delegate the task of sorting email, but not the authority to send a binding legal commitment. A clinician might use a system to identify patterns for review, but not permit it to make an unappealable decision about a patient’s care. A public agency might use automation to organise applications, but not to exclude someone from an essential service without accountable human judgment. A worker might use an AI assistant to prepare a report, but not surrender responsibility for the claims presented in their name.
These examples illustrate a basic rule: delegation must remain proportionate to the consequence of the action. The greater the effect on a person’s rights, livelihood, health, liberty, property, identity or access to essentials, the stronger the requirement for human authority, explanation and review.
The problem begins when systems are given broad permissions under vague language. “Improve efficiency,” “reduce risk,” “personalise the user experience” and “streamline decision-making” can conceal an extraordinary expansion of authority. A system may begin as a tool for advice and gradually become the mechanism that determines who receives an opportunity, who is flagged as suspicious, who is assigned work, whose request is delayed or whose access is restricted.
People may not notice the change because each individual expansion seems minor. A recommendation becomes a default. A default becomes a required workflow. A required workflow becomes a policy. A policy becomes an unchallengeable outcome.
By the time the decision is visibly captured, the transfer of authority has already taken place.
The danger of expanding permissions
Permissions are not merely technical settings. They are grants of practical power. A system that can view information has one kind of power. A system that can alter records has another. A system that can send messages, approve transactions, set priorities, impose restrictions, trigger enforcement or connect to other systems has greater power still.
Every new permission should therefore be treated as a new ethical question. What can this system now do? Who is affected? What mistake, bias, manipulation or misuse becomes possible? What human authority remains above the system? And how can the permission be withdrawn if circumstances change?
In practice, permissions often expand because convenience creates confidence. A system performs one task well, so it is given another. It produces plausible outputs, so review is reduced. It saves costs, so its scope grows. Eventually, the institution no longer asks whether the system should decide; it asks only how quickly the system can be connected to the next workflow.
This is how efficiency becomes an argument for authority. The system appears competent because it acts quickly. The organisation appears modern because it delegates widely. The people affected appear to be receiving a consistent service. But speed and consistency are not substitutes for legitimacy. A fast decision can still be unjust. A consistent rule can still be cruel. An efficient system can still exercise power that no system should possess.
Human authority cannot be preserved merely by placing a person at the beginning of a process and another person at the end. It must exist throughout the chain: in the decision to delegate, in the limits placed on the system, in the monitoring of its conduct and in the ability to interrupt it.
Manufactured agreement
Delegation can also become illegitimate through manufactured agreement. This occurs when a person appears to authorise a system but is not given the practical conditions necessary to understand or refuse the arrangement.
Someone may accept terms because the service is essential. A worker may consent to an AI-managed workflow because declining it risks disciplinary action or unemployment. A customer may allow a system to make decisions because the alternative is delay, exclusion or loss of access. A patient may accept a digital system because it is the only route to support. A citizen may use an automated public service because no meaningful human channel remains.
In each case, the institution may point to a click, signature or continued use as proof of agreement. But formal agreement does not answer the real question: was the person given a meaningful choice?
Meaningful authorization requires material understanding. People should know what task is being delegated, what data will be used, what permissions the system holds, what consequences may follow and how they can challenge or revoke the arrangement. Consent to one function must not be treated as permission for unrelated profiling, targeting, data sharing or autonomous action.
Authorization also requires realistic alternatives. A person cannot be said to have freely delegated a decision if refusing means losing access to an essential service, their basic income, a vital capability or ordinary participation in society. The more consequential the system, the less acceptable it is to make compliance the price of inclusion.
Manufactured agreement is dangerous because it makes control appear voluntary. It allows institutions to say, “The person chose this,” when the institution itself designed the conditions under which the choice was made.
My illustration “The Delegation Throne” work-in-progress. The art represents how unchecked delegation can make human authority merely ceremonial, as systems built to assist gradually assume the power to decide.
Nominal human oversight
“Human in the loop” has become a reassuring phrase. It suggests that a person remains responsible and that automation therefore poses no serious threat to agency. But human oversight can be real, or it can be nominal.
Nominal oversight occurs when the human reviewer lacks the time, information, authority or institutional protection needed to disagree with the system. A reviewer asked to approve hundreds of automated outcomes in a short period is not exercising judgment. A worker who cannot see the reason for a recommendation cannot meaningfully assess it. A manager who is punished for overriding the system does not hold genuine authority. A caseworker who is told that the model is more objective than they are may become the public face of a decision they are not permitted to change.
In these circumstances, the human becomes an accountability shield. They absorb complaints, sign forms and provide the appearance of responsibility while the real authority remains hidden in the system’s design, data, objectives and institutional incentives.
Meaningful oversight requires more. The reviewer must understand the decision well enough to identify when it is wrong or harmful. They must have access to relevant context that the system may have missed. They must have the power to override or pause the process. They must be able to give reasons in their own words. And they must be protected from retaliation when they exercise independent judgment.
Human oversight is not the act of touching a button. It is the retained ability to say no.
Conditions for meaningful authorization
A decision may be delegated only when the authority to delegate is real. The following conditions provide a practical test.
- Specific purpose: The task must be defined clearly. Broad permissions granted for vague goals should not become a blank cheque for new forms of control.
- Proportionate scope: Permissions must correspond to the task and should not exceed what is needed to perform it.
- Material understanding: The person or institution authorising the system must understand its significant functions, consequences, data use and limitations.
- Meaningful choice: Authorization must not be extracted by making refusal practically impossible or disproportionately harmful.
- Human contestability: High-impact outputs must be reviewable by a person with the authority and resources to disagree.
- Traceable responsibility: It must be clear who set the objective, granted the permissions, operates the system and answers for harmful outcomes.
- Time limits and renewal: Delegated authority should not become permanent by inertia. Material changes require renewed authorization.
These conditions do not prevent innovation. They make innovation worthy of trust. An institution that cannot explain why it needs a permission, who will be affected and how the permission can be withdrawn has not earned the authority it seeks.
Revocation must be real
The right to revoke is the final proof that delegation has not become subordination. A person must be able to take back authority when a system acts beyond its mandate, when circumstances change or when the relationship no longer serves their interests.
Revocation should be practical, not merely contractual. A person should not have to surrender their records, identity, data, essential capability or access to basic services in order to withdraw from an automated arrangement. The ability to revoke depends on portability, continuity, alternatives and clear procedures for restoration.
For institutions, revocation means more than an emergency shut-off switch. It means the capacity to suspend a system, investigate harm, correct records, restore access, compensate people where appropriate and redesign or retire a process that cannot be justified. A system that cannot be stopped without institutional collapse has already become too powerful.
For individuals, revocation includes the right to disconnect, to opt out of unnecessary profiling, to seek human review, to correct consequential information and to refuse a technology without being punished for remaining human. These rights are especially important when technology touches the body, mind, identity, work or access to essential opportunities.
Human-Sovereignty Transhumanism provides the central principle: consent before connection, control before integration and accountability before delegation. Technology may extend human capability, but it must not inherit ultimate authority over the person whose life it affects.
Keeping the decision human
The solution is not to reject every automated process or treat all delegation as a threat. The solution is to remember what delegation is for. It is meant to help human beings act more effectively, not to make them spectators in decisions that govern their lives.
When a system assists a person, it should make their choices clearer, not narrower. When it offers a recommendation, it should leave room for judgment. When it acts under permission, that permission should be limited and revocable. When it makes a mistake, a person should be able to understand what happened, challenge the outcome and obtain correction.
The captured decision is not inevitable. It is the result of institutional choices: to grant broad permissions, to conceal objectives, to treat agreement as consent, to equate speed with legitimacy and to call ceremonial review “human control.” Different choices are possible.
A free society does not ask whether humans remain somewhere in the loop. It asks whether they still hold the authority to define the loop, interrupt it and reclaim the decisions that matter.
Delegation serves human freedom only when the authority delegated can still be understood, limited and taken back by the human being it is meant to assist.