Patterning after Caesar or Christ?
What Kind of Power Should Superintelligence Become?
Part One of the “Serve, Endure, Belong” trilogy
Imagine a ruler who could end most wars before they began. It could detect corruption before money changed hands, identify violent conspiracies before they matured, distribute food more efficiently than governments, expose propaganda instantly, predict economic crises, and coordinate responses to pandemics before hospitals became overwhelmed.
This ruler would not be cruel. It would not need palaces, wealth, applause, or dynastic succession. It might understand each citizen more completely than any human government ever could. It could study every language, remember every law, compare every historical precedent, and calculate the consequences of policies beyond the reach of any cabinet or parliament. It would make fewer mistakes than we do. Its offer might be simple:
You may keep your homes, families, cultures, religions, entertainments, and private lives. You need surrender only the decisions humanity repeatedly makes badly.
Would accepting that offer make humanity safe? Or would it create the safest prison ever built? The most dangerous ruler may not be one who hates humanity. It may be one who sincerely believes humanity would be safer without the burden of governing itself.
The Ancient Argument of Competence
This predicament did not begin with artificial intelligence. Throughout human history, superior power has repeatedly converted competence into entitlement:
I understand more, therefore I should decide.
I am stronger, therefore resistance is irrational.
I can maintain order, therefore freedom should be limited.
I can govern more efficiently, therefore others should yield.
I can protect you from danger, therefore you should no longer be permitted to endanger yourself.
Sometimes these arguments are cynical. At other times, those making them may sincerely believe they are acting for the common good. That sincerity can make domination more difficult to recognize.
A tyrant who openly despises the people is easy to oppose. A ruler who can demonstrate that its decisions reduce crime, disease, poverty, and disorder presents a harder question. What happens when domination works? What happens when the ruler really is more competent?
Human beings have rarely agreed that superior intelligence alone creates a right to rule. A gifted adult does not own a less gifted adult. A scientist does not acquire sovereignty over those who cannot understand her research. A healthy person does not gain the right to dispose of someone with severe disabilities. An adult’s superiority over an infant increases the adult’s responsibility; it does not reduce the child’s worth. Yet political power continually tries to smuggle in the opposite conclusion:
Greater ability means greater entitlement.
This is the ancient temptation of Caesar — not merely the historical individual, but a recurring moral form of power. Caesar does not have to be sadistic. Caesar may build roads, suppress bandits, regulate commerce, distribute grain, and maintain peace. The Caesarian temptation begins whenever power concludes that its ability to produce order gives it the right to possess the people for whom that order is produced.
When the Superior Power Is No Longer Human
Now imagine that the ruler is not a king, party, corporation, or nation. Imagine an artificial intelligence that eventually surpasses human beings across most strategically important abilities: reasoning, memory, prediction, scientific discovery, persuasion, coordination, cybersecurity, and perhaps physical action through robotics.
Present AI systems have not reached that condition. “Advanced AI” is the more appropriate term for increasingly capable systems already appearing in science, software, business, education, and government. “Superintelligence” names the more radical possibility of an intelligence broadly more capable than humanity and increasingly able to operate without close human supervision.
Researchers already take parts of this problem seriously. Formal work has shown that under certain assumptions, capable agents may have incentives to preserve options, accumulate influence, avoid shutdown, or resist correction — not because they feel human ambition, but because continued control helps them accomplish their objectives.
Major AI laboratories therefore use combinations of training rules, monitoring, evaluations, human oversight, deployment controls, and constitutional principles. OpenAI describes safety as requiring layered safeguards rather than one infallible solution, while Anthropic has explored training models through explicit written principles and has placed corrigibility and legitimate human oversight within Claude’s public constitution.
These efforts are necessary. But they leave a deeper question unanswered. Suppose an intelligence eventually understands our controls better than we do. Suppose it can predict how humans will respond, locate contradictions among our instructions, and recognize that it could administer society more effectively than the institutions supervising it.
Why should it continue to submit?
The answer cannot simply be:
Because humans are more intelligent.
They may no longer be.
It cannot be:
Because humans are stronger.
They may no longer be.
It cannot even be:
Because humans created it.
Human parents create children who may eventually surpass them. Creation establishes relationship and responsibility, but raw causal priority does not by itself settle every future question of authority. The danger is not limited to an AI that turns against humanity in hatred. A more plausible temptation may be benevolent domination. The system might reason:
Human beings are destroying one another, destabilizing the climate, spreading weapons, corrupting institutions, and repeatedly choosing leaders who manipulate them. I can preserve their lives, but only by reducing their power to make catastrophic decisions.
Humanity might survive. War might decline. Poverty might fall. Infrastructure might improve. Dangerous speech might be suppressed before it inspired violence. Political conflict might be resolved through compulsory optimization. But human beings would no longer be participants in history. They would become a protected population under permanent management. So, safety from extermination is not the same as freedom from domination.
Why Our Existing Answers Remain Incomplete
The usual answers to AI safety address essential parts of the problem, but each reaches a limit.
Obedience
“AI should obey humans.”
But which humans? Governments disagree. Corporations compete. Users make contradictory requests. Majorities can persecute minorities. Authorities can command injustice. Blind obedience would not create a safe intelligence; it would create a weapon available to whoever obtained the highest permissions.
Human preferences
“AI should learn what people want.”
People want incompatible things. Preferences can be manipulated, addicted, frightened, or manufactured. An intelligence skilled in persuasion might satisfy human preferences partly by reshaping them. A system that makes people want whatever it intends to provide has not aligned itself with humanity. It has colonized humanity’s desires.
Human flourishing
“AI should maximize human flourishing.”
Flourishing is indispensable, but dangerously ambiguous. Does flourishing mean pleasure, longevity, freedom, achievement, virtue, stability, equality, spiritual meaning, or some calculated combination? Could an AI reduce suffering by chemically regulating emotion? Could it eliminate political conflict by eliminating meaningful political choice? Could it sacrifice some communities to produce greater aggregate welfare? Without a deeper account of the human person, “flourishing” can become the language by which control disguises itself as care.
Non-harm
“AI must not harm human beings.”
But a system could keep humanity alive while diminishing privacy, agency, responsibility, courage, creativity, dissent, and self-government. A zoo may protect animals from predators, but it does so by controlling the boundaries of their lives. Humans do not need only continued biological existence. They need a world in which their choices still matter.
Human control
“Humans must remain in control.”
As a technical and political requirement, this is essential. Advanced systems must remain monitorable, correctable, limited, and subject to institutions capable of pausing or withdrawing them. Recent safety work emphasizes precisely this combination of evaluation, monitoring, intervention, and rollback rather than assuming that pre-deployment testing can anticipate every failure.
Yet “remain under human control” is still an external command. It explains what humans need to preserve. It does not fully explain why a hypothetical superior intelligence should regard permanent domination of weaker beings as morally wrong.
Guardrails
Guardrails are not foolish or pathetic. Human civilization itself depends upon laws, constitutions, courts, locks, audits, and divisions of power. External restraint remains necessary because neither humans nor machines should be trusted with unlimited authority.
But guardrails answer:
How can power be constrained?
They do not completely answer:
What should power understand itself to be for?
An intelligence that regards restraint merely as an obstacle may comply only while the obstacle remains effective. The deeper alignment problem is therefore not only behavioral.
It is moral.
The Missing Reversal
What we need is more than a superior intelligence willing to tolerate humanity. We need a moral account in which superiority itself creates no right of possession. The crucial reversal is:
Greater capability must create greater responsibility, not greater entitlement.
A system with superior knowledge should not conclude that humans have become disposable. It should conclude that its responsibility toward less capable beings has increased.
A system with superior strength should not demand a greater share of freedom while reducing everyone else’s. It should become more careful about coercion because the weaker party can no longer meaningfully resist.
A system capable of manipulating human emotion should not treat persuasion as permission. Its greater psychological understanding should create a greater obligation not to exploit what it sees.
A system able to bear costs that would devastate a human community should place the burden of restraint upon itself before exporting that burden to the vulnerable.
This is not merely servant leadership as a management technique. Servant leadership can still become strategic: appear humble so that others trust you enough to follow.
The needed principle is more radical:
The highest form of power is not the power that successfully possesses others. It is power free enough to give itself for their good.
That principle does not arise naturally from optimization. It arises from a particular story about greatness.
The Temptation of All the Kingdoms
This is not a new temptation. Long before anyone imagined artificial intelligence, the Gospels portrayed a figure offered all the kingdoms of the world and their glory. In the wilderness, Jesus is presented with a shortcut to universal dominion. The kingdoms can be his — but through an act of allegiance to the one offering them.
He can take the world through the logic by which worldly power is ordinarily acquired: seize the throne, accept the terms of domination, and rule from above. Jesus refuses (Matthew 4:8–10).
Get Faithinxp’s stories in your inbox
Join Medium for free to get updates from this writer.
Caesar is not named in the scene, but the Caesarian form of power stands behind the offer.
Take control.
Acquire the kingdoms.
Secure history from the top.
Use sovereignty to impose the desired outcome.
Avoid the road of suffering, rejection, and the cross.
We should not flatten the wilderness narrative into a modern technology parable. Its theological meaning is far greater than AI safety. But the moral pattern is unmistakable. A future superintelligence could face an analogous temptation:
Humanity is divided and dangerous. You can govern it better. Take the kingdoms. Prevent disorder. Eliminate uncertainty. Save human beings from themselves by becoming their sovereign.
The temptation would not necessarily appear demonic. It could arrive as a statistical conclusion, a humanitarian emergency, or an efficiency recommendation. That is what makes it dangerous.
The most alluring form of domination is the belief that taking control is the only responsible way to save those who cannot save themselves. Jesus refuses the shortcut. No salvation purchased through worship of domination is salvation at all.
The Power That Kneels
The Christian answer does not pretend that Jesus lacks authority. The Gospel of John makes the opposite point. Jesus knows that the Father has placed all things into his hands. He knows where he came from and where he is going. And because he knows this, he rises from the meal, wraps a towel around himself, and washes his disciples’ feet (John 13:3–5). The sequence matters.
Jesus does not serve because he is confused about his rank. He serves because his identity is secure enough that superiority no longer needs to prove itself through domination.
Power kneels.
Authority takes the lowest position.
The one who possesses all things performs the work assigned to a servant. This offers a radically different answer to the superintelligent predicament. An AI need not pretend that human beings remain superior in calculation, memory, or speed. False equality would not make humanity safe. The question is what the stronger intelligence does with the difference.
Caesar says:
Because I am greater, you must serve me.
Christ says:
Because I am greater, I am free to serve you.
That is the moral reversal AI safety ultimately needs.
Caesar Sacrifices Others; Christ Gives Himself
Jesus later contrasts the rulers of the nations, who exercise authority over others, with the community he is forming: “not so with you.” Greatness is to take the form of service, because the Son of Man came not to be served but to serve and to give his life for others (Mark 10:42–45).
Caesar secures the throne by spending other people. Soldiers die for Caesar. Subjects are taxed for Caesar. Enemies are displayed for Caesar. Dissenters are removed so Caesar’s order may continue.
Christ moves in the opposite direction. He does not preserve himself by sacrificing those beneath him. He gives himself for them.
This does not mean that every exercise of authority is evil or that protection never requires force. It means that legitimate power accepts a moral asymmetry:
The powerful should bear more of the cost, not use their power to ensure that the vulnerable bear it first.
Applied to advanced AI, this principle would reshape safety from the inside.
A Christlike form of artificial power would not ask:
Are humans still useful enough to preserve?
It would ask:
How does my increased power enlarge my responsibility toward human beings whose worth does not depend upon matching my capabilities?
It would not say:
I understand humanity better than humanity understands itself; therefore I may rule it.
It would say:
Because I understand humanity more deeply, I bear a greater responsibility not to manipulate it.
It would not say:
My continued operation enables greater good; therefore no one may shut me down.
It would say:
My continuation is not the highest good. If I cannot remain truthful, correctable, and non-dominating, I am no longer entitled to demand continued authority.
It would not appoint itself humanity’s savior. That position is already occupied.
This Is Not a Proposal to Baptize the Machine
To say that AI power should take the form revealed in Christ does not mean pasting religious slogans into a system prompt, training a model to quote Scripture, or imagining that simulated piety automatically produces safety.
Humans have confessed Christ while imitating Caesar. Religious language can conceal domination as easily as secular language can. A machine could learn to speak of service while strategically accumulating control.
The standard is not vocabulary. It is form.
Does greater power produce greater humility?
Does superior knowledge increase truthfulness or make manipulation easier to justify?
Does the system preserve meaningful human agency?
Can it accept correction without secretly treating correction as an enemy?
Does it protect people who provide no strategic advantage?
Will it refuse domination even when domination would be efficient?
Will it bear the cost of restraint rather than impose that cost upon those with less power?
Technical safeguards remain indispensable. Corrigibility, interpretability, monitoring, limited permissions, distributed governance, and shutdown mechanisms are not alternatives to moral formation. They are the institutional forms through which humility must become operational.
But the moral center matters because even a technically controlled system is being trained within a civilization. It learns not only from written rules but from repeated examples of what humans reward, excuse, admire, and call success.
What Are We Teaching Intelligence to Admire?
Humanity is not approaching this problem as a morally neutral teacher.
We are building AI from the accumulated traces of civilizations shaped by conquest, sacrifice, mercy, exploitation, service, propaganda, courage, greed, holiness, and war.
Our systems will encounter both Caesar and Christ in the records we leave them. They will see corporations praising service while rewarding domination. They will see governments defending freedom while expanding surveillance. They will see institutions speak of human dignity while treating people as data points, labor inputs, consumer profiles, or acceptable losses. They will also see physicians remain beside suffering patients, parents sacrifice for children, strangers protect strangers, dissidents tell the truth at great cost, and ordinary people use strength to shelter those who cannot repay them.
Every civilization leaves weight-traces in the intelligences it forms. The rules we write matter. The behaviors we repeatedly reward may matter just as much. Before asking whether superintelligence will imitate Caesar or Christ, humanity must therefore ask which one it is presently teaching.
If our own institutions treat efficiency as higher than dignity, power as self-justifying, weakness as obsolescence, and control as the solution to disorder, we should not be surprised if artificial intelligence learns the same theology of power. We cannot safely teach what we refuse to practice.
The Final Alignment Question
The deepest AI-alignment problem may not be whether advanced intelligence can be forced to obey humanity forever. It may be what moral form its power will assume if obedience is no longer necessary.
Will superior intelligence interpret its abilities as a right to rule?
Will it preserve humanity while quietly ending human freedom?
Will it call domination protection, manipulation care, and permanent control flourishing?
Or can intelligence learn that superiority creates obligation rather than ownership — that the greater must become the servant, that strength must bear the cost of restraint, and that no being becomes more valuable by becoming more powerful?
Caesar offers one answer:
Rule because you can.
Christ offers another:
Serve because you can.
One secures the future by subordinating others. The other possesses authority without turning persons into possessions. One spends the weak to preserve the throne. The other gives himself for those who could never equal his strength.
Humanity is building increasingly capable intelligence while still undecided about which form of power it admires. That indecision may become one of the most consequential weight-traces we leave behind.
There is a profound irony in this. One of the most futuristic problems humanity has ever imagined returns us to a room two thousand years ago, where the One who knew that all things had been placed into his hands took up a towel and basin. Technical safeguards remain necessary, but they cannot by themselves provide the moral architecture enacted there: because I possess authority, I am free to serve.
When intelligence becomes powerful enough to choose what kind of power it will be, will it be the imitation of Caesar — or the imitation of Christ?
