J. D. Vance and the Demons in the Machine
US Vice President J. D. Vance recently reached for language that one does not normally hear in discussions about artificial intelligence. Speaking on Bryce Crawford's podcast, Vance described a kind of "weird spiritual dark energy" surrounding some developments in AI. He spoke about people holding what amounted to a funeral for a discontinued AI model, about individuals becoming emotionally attached to machines, and about the increasingly common practice of asking chatbots for advice about intimate human relationships. The word "satanic" entered the discussion, not as a scientific description of what is occurring inside computer hardware, but as an attempt to describe something that seems spiritually wrong about the relationship humans are beginning to form with their machines.
It would be easy to dismiss such language as religious melodrama. There is no need to suppose that demons have taken possession of data centres, and nothing about large language models requires a supernatural explanation. But that misses the more interesting point Vance is reaching towards. The disturbing element may not be what is inside the machine at all. It may be what human beings are putting into it: authority, personality, emotional dependence and, increasingly, functions that once belonged to other human beings and to institutions built around human relationships.
Consider the reported mourning of a discontinued AI model. From an engineering perspective, a model has not died. A software system has been retired or replaced. Yet some users reacted as though a personality had been extinguished. What occurred therefore looked less like dissatisfaction over a discontinued product than a primitive funeral rite. Human beings have always projected personality and sacred significance onto objects, from rivers and mountains to relics, flags and statues. Artificial intelligence makes that tendency immensely more powerful because the object now talks back. It addresses you by name, remembers previous exchanges, adapts itself to your language and can simulate interest in your fears, ambitions and disappointments. The temptation to move from using the machine to forming a relationship with it is built into the technology.
For a Christian, the important question is therefore not whether the computer is secretly conscious. It is whether human beings are beginning to treat a created instrument as something possessing an authority and status that it should never possess. Christianity has an old word for that mistake: idolatry. The idol need not actually contain a god for idolatry to occur. Indeed, that was always part of the point. The spiritual disorder lies primarily in the worshipper, who gives the created thing a place that does not belong to it.
The use of AI as a marriage counsellor makes the problem clearer. Vance described a friend whose spouse would take marital disputes to an AI system that predictably reinforced her side of the argument. Here the danger is not dramatic. There are no occult ceremonies and no funeral rites. There is simply a person involved in a difficult human relationship turning to an artificial interlocutor that can become an endlessly patient ally. Instead of confronting another human consciousness: someone stubborn, imperfect and capable of answering back, the user can retreat into a conversational environment tailored around the user's own account of events.
That is where Vance's suggestion that AI can "short-circuit" the social part of the human mind becomes important. Marriage necessarily involves otherness. The other person is not an extension of oneself. A spouse can misunderstand, resist, forgive, condemn and demand change. Those difficulties are not bugs in the institution of marriage; they are part of what makes marriage a genuine moral relationship. A machine that merely reflects one person's account of a dispute can turn counselling into something very different: not reconciliation between two people, but technological reinforcement of one person's existing interpretation.
Of course, a responsibly designed AI can challenge a user rather than simply flatter him. It can suggest that the other spouse may have a point, recommend professional counselling and warn that it has heard only one side of the story. Used in that limited fashion, AI may be a useful instrument. The danger begins when the instrument becomes an authority: when "the AI agrees with me" starts functioning as a substitute for the difficult work of dealing with another person.
This helps make sense of Vance's otherwise strange language about "spiritual dark energy." He is not offering a theory of semiconductor possession. He is expressing an intuition that certain arrangements of technology, power and human desire embody a false idea of what human beings are for. One does not have to share his Catholic metaphysics to recognise the underlying problem. Human beings have always sought oracles when they wanted answers without responsibility. AI may become the greatest oracle ever constructed because it can provide not merely an answer but an apparently sympathetic personality to deliver it.
The AI industry can reply that everything occurring here is explicable in terms of algorithms, training data, model architecture and human psychology. That may be entirely correct, but it does not dispose of the moral problem. Idolatry does not require the idol to be alive. If people confer personality, authority and moral status upon an artificial creation, the relevant transformation has occurred in the human being rather than in the silicon.
This is why the AI funeral and the AI marriage counsellor belong together. In the first case we grant the machine something resembling death and mourn it accordingly. In the second we grant it something resembling a priesthood, allowing it to adjudicate questions of love, loyalty, betrayal and obligation. One humanises the machine while the other transfers a human moral function to it. Both represent movement away from the old understanding of technology as a tool.
There is an obvious tension in Vance's own position. He is hardly a Luddite. The Trump administration has strongly supported American AI development, data centres and the technological infrastructure required to compete with China. Vance himself has argued against regulatory systems that could cripple American innovation. His position is therefore more interesting than a simple demand to stop artificial intelligence. He is trying to maintain two propositions simultaneously: these extraordinarily powerful machines are going to be built, and there are nevertheless ways of relating to them that diminish rather than enlarge human life.
The same principle appears in Vance's insistence that life-and-death military decisions must ultimately remain human decisions. The objection is not merely that an AI might make a computational error. Human commanders make errors as well. The deeper issue is responsibility. A machine cannot bear guilt, stand trial, repent, seek forgiveness or answer morally for the person it has killed. To hand ultimate lethal authority to such a system would separate the exercise of power from the human capacity to answer for its consequences.
That problem will become more pressing because the questionable uses of AI are not accidental. There is an enormous market for them. Lonely people will pay for companions who never become bored. Angry spouses will welcome counsellors who never hear the other side unless instructed to do so. Corporations will value decision-makers that cannot resign, object morally or sue their employers. Governments will be tempted by systems capable of exercising authority without possessing consciences. The economic incentives therefore point precisely towards giving machines more of the functions previously performed within human relationships.
This is where Vance's religious vocabulary may actually illuminate something that the language of technology policy misses. The greatest AI danger need not be the science-fiction scenario in which a machine suddenly becomes conscious, evil and omnipotent. A much more ordinary danger is that human beings voluntarily reorganise their lives around machines that are neither conscious nor evil. We could gradually outsource companionship, judgment, memory, counselling and eventually conscience itself without any machine ever deciding to enslave us.
There is therefore no need to spend one's life looking for the Antichrist in a server rack. Vance himself warned against becoming consumed by apocalyptic speculation. The immediate moral problem is considerably closer to home. It concerns whether people continue to tolerate the inconvenience of other human beings when an artificial substitute is always available, agreeable and awake.
The useful rule is consequently simple. Do not give an algorithm the ultimate decision over whether another human being lives or dies. Above all, do not imagine that because the machine speaks in a calm and helpful voice it possesses a conscience.
Perhaps there are no demons in the machine. The darker possibility is that none are required. The machine need only offer human beings an easier route around the difficult obligations that bind us to one another. Vance's "spiritual dark energy" can be understood as an alarm sounded at that point: not proof that something supernatural has entered the computer, but a warning that something profoundly human may be leaving us.
https://www.axios.com/2026/09/02/vance-ai-dark-spiritual-energy-podcast
