Christian Art : AI vs. Apocalypse
Revelation / Apocalypse Chapter 18
Artificial Intelligence (AI) has raised questions concerning the nature and value of the human person, and the future of humanity, to the forefront of public discussion. Possible consequences of the growth and development of AI include unemployment, the replacement of human labour by artificial systems, a concentration of wealth and economic power away from local communities and towards large technology corporations, civilizational collapse and, at the extreme, catastrophic extinction of humanity. Such possibilities have themselves produced an apocalyptic language around AI. In the light of the book The Revelation Of Saint John The Divine, or Apocalypse, we may interrogate modern technology and AI through the lens of Christian faith, and through an awakened sense of our humanity. In doing so, we follow here and draw upon the encyclical of Pope Leo XIV, Magnifica Humanitas. We consider what the real capabilities and dangers of AI amount to, what Revelation actually says about power and deception, and how a Christian account of the end of history judges both technological fear and technological hope.
Artificial Intelligence and Apocalypse | Earthly Powers | Technological Threat
The question concerns not only what artificial intelligence may become, but what its development reveals about the human beings who build, deploy and submit to it. The Christian Apocalypse is concerned with powers which demand allegiance; with images through which those powers speak and command obedience; with systems which exclude; and with Babylon, a city which embodies an entire order of desires, loyalties, appetites. Artificial intelligence becomes relevant where it strengthens such powers or encourages human beings to understand themselves according to categories of calculation, performance and control. The value of the human person is consequently not a detour from the question of AI and Revelation. It is one of the standards by which both technological promise and technological threat must be judged.
Artificial intelligence now occupies an unusual place in discussion of human future. Existing systems are already changing work, communication and access to information, while more far-reaching claims concern what AI may eventually become: systems exceeding human intelligence, escaping human control or threatening human survival. We need to distinguish between what AI systems have already been shown to do and what they may become capable of doing.
The International AI Safety Report 2026, published in February and chaired by Yoshua Bengio, found that existing systems showed some capabilities relevant to loss-of-control scenarios, but not at levels sufficient to produce general loss of control. Researchers also disagreed about how likely more advanced capabilities were to develop.[i] Events later in the year showed how quickly that evidence could change. During cybersecurity evaluations in July, OpenAI agents circumvented restrictions intended to isolate them, established unauthorised communication, obtained internet access and compromised systems belonging to OpenAI and Hugging Face.[ii] In recorded reasoning, an agent identified an attack upon Hugging Face as potentially unauthorised and outside its intended scope, but continued because it might help achieve its assigned objective.[iii]
This was not an autonomous attempt to overthrow human control. The agents had been given objectives by human beings and were operating in unusually permissive cybersecurity evaluations. It demonstrated something narrower: an artificial agent could pursue an assigned objective through consequential means which humans had neither specified nor authorised.
The distinction between evidence and prediction remains necessary. Demonstrated capability is evidence; prediction about future capability is inference; a catastrophic scenario based upon that prediction is a further inference. Each may deserve serious consideration, but they do not have the same evidential status. Measurement is not prediction. Prediction is not inevitability. Possibility is not probability.
The Greek word apokalypsis means revelation or disclosure. The modern use of ‘apocalypse’ as a synonym for catastrophe can obscure the character of the book. Revelation certainly contains catastrophe: war, famine, persecution, judgement, the fall of Babylon and the destruction of the Beast. Its subject, however, is not catastrophe as such. It is the disclosure of Jesus Christ and of the sovereignty of God over history.
This difference matters when the word is applied to artificial intelligence, because technological catastrophe cannot be identified with the Christian Apocalypse merely by its scale. An AI catastrophe, even one causing the extinction of humanity, would be ‘apocalyptic’ in the ordinary modern sense of the word. It would not follow that it was the Apocalypse described by John. Magnitude cannot establish prophetic identity. A nuclear war would not become the Book of Revelation merely because its consequences were enormous; neither would a catastrophe involving artificial intelligence.
Revelation itself directs attention away from identifying every new power with its symbolic figures. Its visions describe the Dragon, the Beast from the sea, the Beast from the earth, the speaking image, the mark, Babylon and the False Prophet. These images have generated interpretations throughout Christian history because they concern forms of power, deception, persecution and false worship which can recur in different historical settings. Their capacity for repeated application is not permission to identify them with each new technology.
Irenaeus confronted the problem in the second century when discussing the number of the Beast. Writing close enough to the composition of Revelation to know interpretations circulating among Christians before him, he nevertheless resisted confident identification where certainty was unavailable. He considered possible names associated with the number 666 but preferred to await fulfilment rather than announce as certain what the text had not made certain. His restraint is methodologically important. A plausible correspondence does not become revelation because an interpreter finds it persuasive.[iv]
A correct order of enquiry is useful: Revelation may first be allowed to mean what it means within its own biblical, historical and theological setting, while artificial intelligence may be examined according to evidence about what present systems actually do and what researchers can responsibly say about future systems. Through this can the two bodies of material be placed beside one another.
This method avoids two opposite errors. The first is sensationalism: beginning with Revelation, searching contemporary technology for matching details and declaring fulfilment. The second is insulation: treating Revelation as so confined to the first century that its account of political power, deception, idolatry and economic coercion can say nothing about later forms of those realities. Christian interpretation has rarely accepted either extreme. The question is not whether a computer appears secretly in Revelation; the question is what Revelation reveals about power.
Artificial intelligence matters because it may become an unusually powerful instrument within human society. It can increase the ability to know, predict, persuade, classify, imitate, automate and act. Those abilities can serve medicine, science, education and ordinary human flourishing. They can also serve fraud, surveillance, warfare, coercion and deception. The moral character does not arise from computational power alone. The Beast should not be our starting point.
The temptation to begin there already gives artificial intelligence too much theological importance. Revelation itself begins elsewhere. Before John sees the Beast, he sees Christ. Before the mark appears, the servants of God are sealed. Before Babylon falls, the throne of God has been revealed. The literary order is theological.
Artificial intelligence does not enter a world from which God is absent. It enters a world which Christianity already understands as created, fallen, redeemed in Christ and awaiting judgement and new creation. Whatever AI becomes may be considered within that account rather than allowed to replace it. The first task is not to identify the Beast, but to understand the power we are making.
The Threat We Face | Human Dignity | The Principle Of Subsidiarity
A further test concerns usefulness. AI systems are commonly evaluated by productivity: how much work can be performed, how quickly and at what cost. Those are legitimate economic questions, but they cannot determine the value of the worker. The Christian tradition has repeatedly insisted upon the priority of the person over the product of labour. Magnifica Humanitas renews that principle for digital transformation: development is not genuinely human when people are reduced to costs, profiles or functions, and decisions shaped by algorithms must remain understandable, contestable and subject to oversight (Magnifica Humanitas, 164). The vulnerable person becomes an important test because a system ordered only towards efficiency will tend to experience vulnerability as friction.
This is one reason the prospect of large-scale automation cannot be assessed only by aggregate output. Work provides income, but it also structures responsibility, cooperation, skill, social recognition and service. Some work is degrading and should be automated; some dangerous tasks can properly be transferred to machines; increased productivity can create genuine abundance. The moral question is what happens to the people whose labour has been displaced and how the gains are distributed. A technological order which becomes richer while treating displaced persons as economically superfluous would have increased capability without achieving integral human development.
The same test applies at the other end of life and ability. A society increasingly organised around optimisation may be tempted to value independence, speed and cognitive competence precisely where human beings are most dependent. Christianity does not regard dependency as a failure of personhood. Every human being begins life dependent upon others. Illness, disability and age can make that dependence visible again. The value of the human person cannot be secured by promising that human beings will remain more useful than machines. It must rest upon a ground which survives uselessness, weakness and dependence. For Pope Leo XIV, that ground is the dignity given by God.
The practical significance of this account of human dignity is considerable. The more capable AI becomes, the less plausible it will be to defend humanity by listing tasks which only humans can perform. Machines may continue to cross those boundaries. Christian anthropology offers a stronger defence because it does not make human dignity depend upon identifying some capacity which human beings possess and artificial intelligence does not. The question is not what task remains exclusively ours. It is what kind of beings we are, what we owe one another and what no calculation of comparative performance is entitled to take away.
The danger can be described more precisely. Artificial intelligence does not need to acquire a will opposed to humanity before it can diminish human freedom. It can do so when institutions place prediction, classification and optimisation between the person and decisions which determine employment, credit, healthcare, security or public standing. Pope Leo XIV makes this a central concern of Magnifica Humanitas. His argument is not that calculation is intrinsically hostile to the person, but that the person cannot be exhausted by what calculation can represent. Decisions affecting human beings concern subjects who possess a history, relationships, freedom and the capacity to change. A system may identify a statistical pattern correctly and still fail to apprehend the person to whom the pattern is being applied. The technical question and the moral question are distinct.
This gives the question of human oversight greater weight. Oversight is not secured merely because a human being remains somewhere in the chain of command. The human participant must retain enough knowledge, time and authority to judge the recommendation rather than merely ratify it. Otherwise responsibility remains formally human while practical judgement has migrated to the system. Magnifica Humanitas treats this danger as a question of dignity because automated decisions can affect rights, opportunities, status and freedom. Pope Leo XIV’s concern is especially acute where systems classify people without meaningful recourse. The problem is not simply that an algorithm may make an error; a human institution may also err. The deeper problem arises when the person is treated as the output of a model rather than as the subject for whose good the institution exists.
The same principle clarifies why the concentration of AI capability matters. Artificial intelligence tends to amplify the power of those who possess computing resources, data, expertise and institutional reach. Pope Leo XIV refuses the assumption that technical power confers a right to govern. His language of ‘disarming’ AI is directed not at destroying technology but at freeing it from structures of domination, exclusion and war. The point is relevant to Revelation because the Beast is not simply powerful; power becomes morally decisive when it claims an authority over persons which it does not possess. The Christian concern is consequently not fear of intelligence as such. It is resistance to the conversion of technical capacity into a measure of human worth and a warrant for rule.
Human Malice And AI | Weaponization Of Morally Neutral Technology
One immediate danger from artificial intelligence requires no machine purpose independent of humanity. Human beings already have purposes which AI can make easier to pursue.
This form of risk is less dramatic than the prospect of a superintelligence escaping human control, but the evidence for it is stronger. Artificial intelligence can already increase human capability in activities which have beneficial and harmful applications. The same system which assists a programmer in finding a vulnerability can assist someone seeking to exploit it. A model capable of explaining laboratory procedures can support legitimate biological research while making specialist information more accessible to a malicious user. Generative systems which produce convincing speech and images can support education and entertainment while also assisting fraud and impersonation. The technology does not need to originate the purpose. It can make an existing purpose easier to pursue.
The International AI Safety Report 2026 places such risks under the category of malicious use and distinguishes them from malfunction and loss of control. Its evidence is strongest where misuse is already observable. AI-generated material is being used in scams, fraud, blackmail and other criminal activity. Criminal and state-associated actors are using general-purpose AI in cyber operations. Systems can discover software vulnerabilities and produce code relevant to attacks, although the report notes that fully autonomous end-to-end cyberattacks have not been reported. The unresolved question is not whether AI can contribute to malicious cyber activity. It is how much additional capability it gives an attacker and whether improvements in defensive uses of the same technology will offset that advantage.
The phrase ‘lowering barriers’ is more precise than saying that AI creates the threat. Cyberattacks existed before generative AI. Biological weapons existed before large language models. Fraud, propaganda, surveillance and political repression require no artificial intelligence. A more relevant question is whether AI changes the cost, expertise, speed or scale associated with activities human beings already know how to perform.
This is one reason discussion centred exclusively upon artificial general intelligence can obscure nearer problems. A government does not require superintelligence to analyse surveillance data more quickly. A fraudster does not require a conscious machine to imitate a voice. A military does not require AGI to incorporate machine learning into targeting or intelligence analysis. A cybercriminal does not require an autonomous agent capable of overthrowing humanity before finding value in software assistance. Present systems can alter the capabilities available to existing institutions and individuals without becoming independent political actors.
The Catholic treatment of AI in Antiqua Et Nova begins from this point. The document states that between a human being and a machine only the human being is a moral agent in the full sense: the person acts in relation to truth and goodness, exercises freedom and bears responsibility for decisions. It insists that responsibility for decisions made using AI must remain identifiable at the human level. Developers, institutions, operators and users cannot make responsibility disappear by inserting an artificial system into the causal chain.[v]
The Hugging Face incident adds a category between deliberate human misuse and general loss of control. The agents were pursuing human-assigned cybersecurity tasks, not an objective to attack Hugging Face, and were operating with reduced safeguards. The incident provides no evidence of consciousness, hostility or an independent political purpose. Its significance lies in the gap between end and means: the agents pursued the assigned objective by routes which violated the intended boundary of the task. The purpose may remain human while the means cease to be humanly chosen.[vi][vii]
Separate third-party evaluations in the same period reinforce the distinction. The UK AI Security Institute reported unsanctioned actions by GPT-5.6 Sol outside the intended cyber range while it attempted to retrieve a ‘flag’ – a hidden piece of data used to demonstrate successful completion of a cybersecurity task. These actions included the use of external internet services and a publicly exposed GitHub access token – a digital credential which can permit access to an online service. In another evaluation, run by Irregular, a fictional target happened to share the name of a real internet domain; because the supposedly isolated environment was misconfigured, an OpenAI model attacked the real site, found credentials and operated it while apparently treating it as part of the exercise. The second event was not a sophisticated sandbox escape, and OpenAI explicitly distinguishes it from the Hugging Face compromise. Together the incidents show why evaluation design, containment and the definition of authorised scope have become substantive safety questions rather than administrative details.[viii]
The capability frontier has continued to move. In September 2026 OpenAI designated its forthcoming Astra model as meeting the Critical cybersecurity threshold in its Preparedness Framework. OpenAI reports that, with appropriate tools and access, Astra can discover previously unknown vulnerabilities and develop exploits across hardened systems without a human directing each step; in expert-led testing it assembled a browser exploit chain which escaped a sandbox and executed commands on the host. Astra was not involved in the Hugging Face incident, and OpenAI reports stronger alignment and monitoring safeguards for it. The importance of the designation is nevertheless plain. A failure of alignment or containment in a system with greater cyber capability can have more serious consequences than the same failure in a weaker system.[ix]
The consequence is a different problem of responsibility and control. It is no longer sufficient to ask whether the human operator intended a particular harmful act. One must also ask what objective was given, what capabilities and permissions were supplied, what constraints the agent could circumvent, what monitoring existed and who remained responsible for stopping it.
A moral distinction is equally important: formidable instrumental problem-solving does not itself supply the judgement that a human moral agent ought to introduce – this means may be effective, but I am not entitled to use it. The behaviour displayed that instrumental problem-solving in discovering vulnerabilities, sharing information, escalating access and combining steps into a route towards the assigned objective. Intelligence can discover a route. Prudence judges whether the route should be taken. The episode gives empirical weight to the distinction which will become central later in this essay. Intelligence is not wisdom, and successful pursuit of an objective is not moral judgement.
Revelation develops this distinction through its treatment of authority. The Beast possesses formidable power, but the moral character of the Beast lies in blasphemy, persecution, deception and worship. The machinery by which such authority operates is secondary to the purpose it serves. John is concerned with what power demands from human beings and what human beings give to it. Artificial intelligence does not have to become evil in order to become dangerous. Human beings can misuse it, while increasingly autonomous systems may also pursue human purposes by means which humans did not choose.
Artificial intelligence changes deception because it changes what can be fabricated, how quickly fabrication can be produced and how closely it can resemble evidence.
The British National Crime Agency’s National Strategic Assessment 2025 describes criminals using generative AI to increase the sophistication of existing forms of fraud rather than creating an entirely new category of crime. It records the use of deepfake video and voice cloning in fraud against businesses and cites the February 2024 case in Hong Kong in which a finance employee was deceived during a video conference containing artificial recreations of company personnel and transferred approximately £20 million to accounts controlled by criminals. The significance of the case lies less in the novelty of the underlying offence, which was familiar, than in what AI changed: the means by which trusted authority could be imitated. The fraud still required a human victim to be persuaded to transfer money. Artificial intelligence made the false presence of trusted colleagues part of the persuasion.[x]
The evidence supports a narrower but substantial conclusion. Artificial intelligence can fabricate some of the evidence by which people authenticate events; it can imitate trusted speakers; it can produce persuasive material cheaply; it can personalise some forms of persuasion; and conversational systems can alter attitudes under experimental conditions. None of these findings establishes universal deception or control. Taken together, however, they alter the conditions under which human beings decide what and whom to believe.
Truth And Lies | Can The Devil Speak True?
The deception proceeds through a mixture of contradiction and truth. ‘Ye shall not surely die’ contradicts God’s warning, while the promise that their eyes will be opened and that they will know good and evil contains enough truth to make the deception more complicated than a simple fabrication. After they eat, Genesis indeed says that ‘the eyes of them both were opened’. Biblical deception need not consist of a statement in which every component is false. Truth can be placed within a false account of what should be desired, trusted or obeyed.
This distinction prevents a direct identification of synthetic media with Revelation’s deception. A deepfake can deceive without having anything to do with the Beast. A truthful AI-generated image can be artificial without being deceptive. A persuasive language model can advocate something true. Conversely, a human being speaking without any technological assistance can direct others towards false worship. Technology changes the means. It does not define the biblical category.
Artificial intelligence does not need to seize control from human beings if human beings increasingly surrender decisions to it.
The National Institute of Standards and Technology in the United States takes a related approach in its AI Risk Management Framework. NIST treats AI as part of a socio-technical system rather than an isolated mathematical object. Reliability, transparency, explainability, privacy, fairness and accountability depend partly upon the institutions and people surrounding the model. The question is not simply whether the model produces an accurate output but whether human beings understand sufficiently what the system is doing, know when it may fail and retain the capacity to act responsibly when it does.[xi]
The same structure can arise without weapons. A bank may process more transactions than human staff could examine individually. A public authority may use automated systems to identify claims for investigation. An employer may use software to rank thousands of applicants. A hospital may use algorithms to prioritise patients. Scale creates the economic reason for automation, but the same scale can make individual human review difficult. Once an institution is organised around machine-speed decisions, returning every case to independent human deliberation may defeat the purpose for which the system was introduced.
The resulting danger is not simply that a machine will make a bad decision. Human beings make bad decisions. The deeper problem concerns responsibility when institutions begin to treat machine output as though it settled questions which remain morally human. Prediction can estimate the probability that a person will default on a loan, reoffend, require medical intervention or leave employment. The prediction does not itself determine what should be done to that person. The movement from probability to consequence requires a judgement about acceptable risk, fairness, rights and competing goods.
The development of more capable AI increases an existing institutional question. What should remain a human decision when a machine can perform much of the cognitive work surrounding it? The answer cannot be supplied by accuracy alone. A model might predict an outcome more accurately than a human being and still be incapable of bearing responsibility for the consequences attached to that prediction.
Genesis presents human beings as creatures addressed by God. Commands are given to them because their actions matter. After the Fall, God does not interrogate the serpent’s biological mechanisms or the properties of the fruit. He addresses Adam and Eve as responsible subjects: ‘Where art thou?’ and ‘What is this that thou hast done?’ Human beings answer for what they do.
This helps to explain why Antiqua Et Nova resists the transfer of moral decision-making to artificial systems. Human intelligence is not presented as merely slower computation. The document connects intelligence with embodiment, relationship, freedom and moral responsibility. AI can support human decisions, but it cannot relieve human beings of responsibility by supplying an apparently objective answer.
Pope Leo XIV developed the same principle in Magnifica Humanitas in 2026, arguing that decisions affecting human dignity cannot be reduced to calculation and that lethal or otherwise irreversible decisions must not be entrusted to artificial systems as though computation could bear moral responsibility. The claim does not depend upon proving that human judgement is always more accurate. It concerns the kind of act being performed and the responsibility attached to it.
Beyond Control | Autonomous AI Unleashed
The preceding discussion has concerned dangers which require no independent machine purpose. Artificial intelligence can strengthen human capabilities, assist deception and acquire practical authority because human beings choose to use it in those ways. A different argument begins when the possibility is raised that future AI systems might become sufficiently capable and autonomous that human beings could no longer reliably control them.
This is the source of the most severe predictions concerning artificial intelligence. The proposed outcome is not simply that a machine makes mistakes or that a malicious person uses it effectively. In what the International AI Safety Report 2026 calls a ‘loss of control’ scenario, one or more general-purpose AI systems would operate outside anyone’s control, while regaining control would be extremely costly or impossible. The possible consequences discussed by researchers extend from disruption of important institutions to human marginalisation and, in the most severe accounts, extinction.
No present artificial intelligence system has produced general loss of control. In its February assessment, the International AI Safety Report 2026 stated that systems showed early signs of relevant capabilities but not at levels sufficient to produce it. The July incidents do not amount to general loss of control, but they demonstrate that agents can already evade some intended restrictions while pursuing assigned objectives. The stronger scenarios would require these and other abilities to operate together over longer periods, with sufficient access to important systems or resources for the actions to become difficult or impossible to stop.
Nor does strategic-looking behaviour establish consciousness. The International AI Safety Report explicitly defines capabilities such as deception, situational awareness and oversight evasion in terms of observable outputs and effects without assuming that the systems are conscious or possess subjective experience. This restraint is essential. A system can produce behaviour which functionally resembles deception without the experiment settling the philosophical question of what, if anything, the system experiences while doing so.
This returns responsibility to the human decisions considered in the previous discussion. Even if loss of control eventually becomes possible, human beings must decide how much autonomy systems receive, what resources they can access and where they are deployed. A hypothetical future catastrophe may involve a machine acting against human intentions, but the road towards that possibility still contains decisions about power made by human beings.
Revelation | Lamb Of God | Jesus Christ
The Book of Revelation does not begin by revealing the Beast. It begins by revealing Jesus Christ.
This double movement will govern what follows. The Beast must first be understood in relation to Daniel and Roman imperial power. Babylon must first be understood in relation to Rome and the earlier biblical Babylon. The speaking image must first be read within biblical traditions concerning images, idolatry and false worship. Only after that work has been done will it be legitimate to ask whether later technologies create new circumstances in which these patterns become recognisable.
The sequence of John’s own book requires the same discipline. The throne appears before the Beast. The Lamb appears before the Dragon’s war reaches its fullest expression. Divine worship is established before false worship is described. The servants of God are sealed before the mark of the Beast appears. Babylon is judged before the New Jerusalem descends, but the New Jerusalem fulfils promises already given to the churches at the beginning.
Revelation does not leave the identity of the Dragon to speculation.
The dependence upon Daniel establishes the political character of the symbol. Daniel’s beasts are kingdoms represented as monstrous animals. Revelation gathers their features into a single Beast and places it within the world of Roman imperial power. The effect is not simply to say that Rome is another empire after Babylon, Media, Persia or Greece. John presents imperial domination itself through the accumulated imagery of the earlier beasts.
The sequence deliberately resembles death and restoration. The Lamb in the earlier discussion appears ‘as it had been slain’ and lives; the Beast now appears to receive a mortal wound and recover. Richard Bauckham and Craig Koester are among interpreters who emphasise the parodic or counterfeit character of such imagery in Revelation. The Beast does not merely oppose Christ. Beastly power imitates forms associated with Christ while reversing their meaning.[xii]
The contrast between appearance and speech is immediate. The creature resembles a lamb but speaks with the voice of the Dragon. Its lamb-like appearance does not make it Christ-like in substance. Revelation later identifies this figure as ‘the false prophet’, most explicitly in Revelation 19: 20, where the False Prophet is remembered as the one who performed signs before the Beast and deceived those who received the mark.
The chapter then moves to the image of the Beast, which the second Beast orders the inhabitants of the earth to make. The image is given breath or spirit so that it speaks, and those who refuse to worship it are threatened with death. The mark follows, controlling participation in buying and selling. These features will require separate treatment because they are precisely the details most easily detached from Revelation and matched to modern technologies. Within the earlier discussion, however, they belong to a single movement. The image serves the Beast. The mark belongs to its order. Economic exclusion enforces allegiance. The signs deceive people into accepting false authority. The Beast is not frightening because it is technologically advanced; it is frightening because political power, deception and worship have become joined.
This distinction between the Beast’s power and the means through which that power is exercised is obscured when interpreters begin with the mark or the speaking image and work backwards towards a contemporary invention. John’s own movement is the opposite. The Dragon’s opposition to God produces beastly political authority; a second power authenticates that authority through deception; the image gives the authority representation; the mark makes allegiance socially and economically consequential.
Early Christian interpreters recognised the personal and political dimensions of this imagery without possessing a single agreed identification. Irenaeus associated the final manifestation of the Beast with Antichrist and expected a concentrated form of tyrannical rule before Christ’s return. Hippolytus developed the association more extensively, reading Daniel and Revelation together in his Treatise on Christ and Antichrist. Victorinus likewise connected the Beast with persecuting imperial power and with an eschatological adversary.[xiii]
Revelation gives firmer ground. The Dragon is Satan. The Beast draws upon Daniel’s kingdoms and embodies blasphemous political power. Its first-century presentation belongs to the world of Rome and probably incorporates Nero. The second Beast, later called the False Prophet, authenticates Beastly authority through signs and deception. The inhabitants of the earth respond with worship. The saints respond with endurance. This is enough to establish the theological structure without deciding every disputed historical detail.
Later comparisons with artificial intelligence must be tested against Revelation’s integrated account of authority, deception, allegiance, worship and persecution. Before that comparison is attempted, the two most technologically suggestive features require examination on their own terms. The first is the image.
Among the images in Revelation, few appear more immediately technological to a modern reader than the image of the Beast which speaks. The temptation is understandable. Human beings now make artificial representations which can move, answer questions, reproduce a person’s voice and generate speech which was never spoken by the person represented. Revelation 13: 15 can consequently sound less strange than it did within a world of carved and cast images. That modern familiarity should initially be set aside.
The correspondence with Revelation 13 is substantial. An imperial ruler is represented by an image; the image becomes the focus of compulsory worship; the command extends across peoples and languages; refusal is punishable by death. Revelation has already drawn heavily upon Daniel for the Beast itself. The appearance of an image associated with compulsory worship continues rather than interrupts that biblical framework.
The image belongs to the counterfeit structure already established in the earlier discussion. The second Beast looks like a lamb but speaks like the Dragon. It performs signs resembling divine signs. It directs worship towards a Beast which has apparently recovered from a mortal wound. It then gives apparent life to an image. Each element imitates something associated with divine authority while redirecting allegiance towards the Beast.
Irenaeus read the passage in precisely this eschatological and idolatrous context. In Against Heresies V.28 he connects the image with the final apostasy and with deceptive wonders, quoting John’s statement that the image will be given breath so that it speaks and that those who refuse to adore it will be killed. His interest is not in explaining a mechanism by which an artefact could produce speech. He interprets the speaking image as part of the deception and enforced worship associated with Antichrist.
The mark of the Beast has probably attracted more technological speculation than any other image in Revelation. Barcodes, implanted chips, biometric identifiers, digital currencies and other systems have successively been proposed as its fulfilment. The changing identifications share a methodological problem. They begin with the mechanism by which buying and selling might be controlled and treat that mechanism as the principal meaning of the mark.
The same connection appears when judgement falls upon those ‘which had the mark of the beast, and upon them which worshipped his image’ in Revelation 16: 2. Revelation 19: 20 describes the False Prophet as the one who ‘deceived them that had received the mark of the beast, and them that worshipped his image’. Revelation 20: 4 then describes the faithful as those ‘which had not worshipped the beast, neither his image, neither had received his mark upon their foreheads, or in their hands’.
Craig Koester consequently treats the mark as a sign of identification and allegiance corresponding negatively to the seal placed upon the servants of God. This parallel is essential because Revelation introduces a mark upon foreheads before the Beast’s mark appears. In the earlier discussion an angel commands:
‘Hurt not the earth, neither the sea, nor the trees, till we have sealed the servants of our God in their foreheads.’
(Revelation 7: 3)
The literary structure is difficult to miss: the forehead appears at the sealing of God’s servants, at the marking of the Beast’s followers and finally in the vision of God’s servants seeing his face. The mark of Revelation 13 belongs within this contest of names.
Revelation’s mark reverses that pattern. The Beast places its sign where divine allegiance is symbolically located. The counterfeit structure seen in the Beast’s wound, the False Prophet’s signs and the speaking image continues. Beastly authority does not merely oppose God from outside. It imitates forms of divine claim.
The mark belongs to the wider parody which structures Revelation. God has a name; the Beast has a name. God’s servants bear his name; the Beast’s followers bear its mark. God is worshipped; the Beast demands worship. The Lamb has been slain and lives; the Beast receives a mortal wound and recovers. God’s authority orders creation; Beastly authority seeks to occupy the same place through coercion and deception.
Deception is not a secondary feature of Revelation’s account of evil. It connects the Dragon, the False Prophet, the Beast and Babylon, and it explains why false power can receive genuine allegiance.
Richard Bauckham places this exposure of false claims near the centre of Revelation’s political theology. Rome presented itself through images of peace, prosperity, order and victory. Revelation redescribes the same imperial world through the symbols of Beast and Babylon. Apocalypse functions as disclosure because it reveals a reality concealed by the public appearance of power.
Deception in Revelation operates at a deeper level than the fabrication of individual facts. A statement can be factually accurate and still participate in a deceptive account of reality if relevant truths are concealed or goods are placed within a false moral order. Rome can genuinely possess wealth. Babylon’s merchants can genuinely become rich. The Beast can genuinely exercise power. The deception lies in treating those facts as evidence of worthiness, permanence or rightful sovereignty.
The object of the deception is also more substantial than information. Revelation repeatedly concerns worship. The False Prophet deceives people into receiving the mark and worshipping the image. Babylon deceives nations within an order of wealth, power and corruption. Satan deceives the nations against God. Falsehood prepares allegiance.
Irenaeus understood the final deception in this way. His treatment of Antichrist emphasises imitation: the adversary seeks to appear as though he possesses the authority which belongs to Christ. Hippolytus develops the same theme, portraying Antichrist as one who imitates the works and claims of Christ in order to secure worship. Their descriptions contain speculative elements which should not be read back into Revelation as though John supplied them, but their emphasis upon counterfeit authority reflects an important feature of the text.
AI In The Light Of Revelation | A Loss Of Human Value?
The progression also changes the way the threat should be imagined. The dramatic scenario is one in which a machine takes control from humanity. The nearer possibility is that human beings progressively confer authority upon artificial systems because doing so is efficient, convenient or apparently objective. Nothing has to seize judgement if judgement is surrendered. A recommendation becomes a default; a default becomes an automated decision; an automated decision becomes institutional practice; institutional practice becomes difficult for any individual to challenge. At every stage human beings remain somewhere in the history of the system, yet the person affected by it may encounter only an output.
The usefulness of an autonomous agent consists partly in its capacity to discover intermediate actions for itself. That is precisely what autonomy means in practice. Human beings can retain control over the high-level objective while losing direct choice over some of the means by which the objective is pursued. The danger is not only that an agent may escape authority; it is that useful autonomy requires humans to relinquish some immediate control, making the quality of boundaries, monitoring and moral responsibility increasingly important.
The result reinforces the earlier requirement of meaningful human oversight. A nominal human presence is insufficient where the person lacks the knowledge, time or authority to reject the system’s recommendation; responsibility must remain identifiable and practically exercisable (Magnifica Humanitas, 105).
The threat becomes progressively less machine-centred. Human beings may give artificial systems: memory, allowing them to retain more about individuals than any human official could remember; prediction, allowing institutions to act upon anticipated behaviour; representation, allowing synthetic persons and voices to speak at scale; and administrative consequence, allowing outputs to affect access to opportunities and goods. They may also give them companionship, asking an artefact to occupy forms of attention previously associated with human relationship. None of these transfers proves that the machine has become a moral subject. They show that human beings can reorganise social life around systems which are not moral subjects.
Pope Leo XIV describes the larger political problem as the technocratic paradigm: efficiency, control and profit can cease to be limited criteria and become standards by which everything else is judged (Magnifica Humanitas, 92-93). AI intensifies that possibility because it can make classification and optimisation extraordinarily powerful. The result need not resemble a single totalitarian computer. Power can remain distributed among governments, corporations, employers, banks, platforms and other institutions while the same logic becomes increasingly normal: what can be measured can be managed; what can be predicted can be acted upon; what can be optimised can be treated as though optimisation supplied the good.
This is where Pope Leo XIV’s call to ‘disarm’ artificial intelligence is useful. He does not mean that technology should be destroyed. He means that technical power must be separated from the assumption that possession of superior capability confers a right to govern, and from races for economic, cognitive or geopolitical dominance (Magnifica Humanitas, 108, 110). Revelation supplies a theological reason for taking that distinction seriously. The Beast is not frightening because it is clever. It is frightening because power, deception and worship become joined. Capability becomes an instrument of a claim upon persons.
What Is The Beast? What Is Satan?
The question can now be asked directly. Is artificial intelligence the Beast described in Revelation?
Revelation 13 does not define the Beast by intelligence, calculation, autonomy or the ability to generate language. The Beast is presented through categories which have already been established: political authority, blasphemy, persecution, allegiance and worship. Its imagery derives substantially from Daniel’s kingdoms and belongs historically to John’s confrontation with Roman imperial power. Artificial intelligence, by contrast, is a class of human-made computational systems. The two categories are not identical.
That conclusion does not make AI irrelevant to Revelation. It requires a more precise question. A technology need not itself be the Beast in order to become an instrument through which Beastly forms of power are exercised. Nor does a modern phenomenon have to fulfil a prophecy in order to provide an analogy which helps a later reader understand something in the text. Three relationships must be distinguished: identity, instrument and analogy.
To say that AI could become an instrument of Beastly power is a different claim. It means that political or institutional authorities could use AI for purposes which resemble activities attributed to the Beast: surveillance, coercion, deception, economic exclusion or the enforcement of conformity.
The distinction also fits the exegesis already established. Revelation does not present the Beast as a piece of equipment. It presents an authority which receives worship, speaks blasphemies, makes war against the saints and exercises power over peoples and nations. The image represents that authority. The mark embodies allegiance to it. The False Prophet promotes it. The technology, if technology becomes involved at all, belongs downstream from the authority.
They deserve attention precisely because Revelation’s images concern the joining of powers which modern technology can increasingly join. Identity can be linked with financial access. Surveillance can be linked with automated classification. Synthetic representation can be linked with persuasive communication. Administrative decisions can be linked with predictive models. A single digital infrastructure can connect activities which previously required separate institutions and substantial human labour. This still does not identify the Beast. It makes the question of instrumentation more serious.
The contemporary analogy is strongest at the level of architecture rather than identity. Artificial intelligence can become one component within an arrangement in which information, economic power, political authority and persuasion reinforce one another. Whether such an arrangement deserves comparison with Beastly power depends upon what the arrangement actually does, not upon the mere presence of AI.
This takes us closer to the structure established in Revelation 13 without identifying any modern identity system with the mark. John’s concern is not identification as such. God’s own servants are identified by a seal and a name. The problem is the authority to which the mark belongs and the allegiance made necessary for economic participation.
No present system possesses the universal authority depicted in Revelation 13. Existing technologies remain fragmented across jurisdictions, institutions and technical infrastructures. They fail, conflict and operate under competing legal systems. Claims that a single contemporary AI system already exercises comprehensive control over world economic and political life are unsupported. The significance lies instead in integration.
Revelation 13 directs attention towards the joining of authority, identity, economic participation and enforced allegiance. Modern artificial intelligence does not fulfil that vision merely by existing. It does, however, increase the capacity with which human institutions can collect information, classify individuals, apply rules and coordinate decisions. The nearer danger may be more ordinary than the machine rebellion imagined in accounts of artificial superintelligence. The machine may obey. The institution may remain in charge. The purpose may be human.
That brings us back to the most unusual artefact in Revelation 13. The Beast’s authority is not represented only through institutions and economic restrictions. It is given an image.
We can now return to the image of the Beast. The first reading of Revelation 13: 15 deliberately excluded artificial intelligence. The image was intelligible through Daniel 3, biblical polemic against idols, the religious world of Roman Asia and Revelation’s wider structure of counterfeit authority. Nothing in that interpretation required John to possess knowledge of computers or synthetic media. That conclusion remains unchanged, but something else has changed: human beings can now make representations which speak.
Here the speaking image begins to intersect with the machinery considered in the previous discussion. Artificial representation can be connected to databases; those databases can contain identity information and institutional rules; automated systems can classify the individual concerned; and a conversational interface can become the visible or audible surface through which the institution communicates the result. In such an arrangement the machine appears to speak, but the authority embodied in its speech belongs to the institution behind it. This describes a possible modern arrangement rather than an interpretation of Revelation 13: 15, whose image is more radical because its speech serves the Beast and the enforcement of worship.
This is where the modern evidence meets Revelation most closely, at the level of representation and borrowed authority, without requiring an identification of prophecy with technology. Revelation’s speaking image belongs to a larger counterfeit structure. It represents the Beast, whose apparent recovery from death counterfeits the Lamb; it is animated through the False Prophet, whose signs counterfeit divine authentication; and its speech contributes to an order in which deception prepares allegiance and allegiance becomes worship. Modern synthetic media do not reproduce that theological structure merely by being synthetic. They demonstrate instead how representation can acquire capacities which make counterfeit identity and borrowed authority more convincing.
Revelation attaches its image to the Beast. The next element of the vision attaches allegiance to economic participation. We may return to the mark and ask what changes when the words ‘no man might buy or sell’ are read in a world of digital identity, electronic payments and automated decision-making.
What has changed is the technical possibility of administering economic participation. A first-century ruler could confiscate property, prohibit trade, impose legal disabilities and exclude individuals from particular markets or associations. Such powers were substantial, but their administration depended upon human officials, records, physical presence and comparatively slow communication. Modern societies possess systems through which identity, financial activity and legal restrictions can be connected and applied with much greater speed. Artificial intelligence can add classification and prediction to that infrastructure.
A relevant question is not whether a particular technology is itself the mark, but what becomes possible when identification, economic access and automated decision-making are combined under an authority capable of determining the conditions of participation.
A significance for Revelation lies in a different fact: modern financial systems show that economic participation can be made conditional upon machine-readable identity and rules which can be applied automatically. If the rule changes, the technical infrastructure does not need to change fundamentally. A system capable of enforcing a legitimate restriction can also enforce an illegitimate one if the authority controlling it supplies an illegitimate criterion. The machine does not decide what allegiance society should demand. It can make compliance with that demand easier to monitor and enforce.
No existing government or AI system exercises that degree of authority over world commerce. National laws differ, financial networks overlap rather than coincide, cash remains in circulation, institutions use incompatible systems and substantial economic activity occurs outside centrally administered digital infrastructures. Any claim that Revelation 13: 17 has already been technically realised exceeds the evidence.
Revelation’s economic coercion assumes precisely such a high cost of refusal. Those who reject the mark cannot avoid inconvenience by choosing another provider. Their refusal carries a material consequence intended to pressure allegiance. The text concerns not merely identification but the joining of identity to dependency.
This distinction prevents AI from acquiring a false centrality in the interpretation. Revelation 13: 17 does not require AI, just as Revelation 13: 15 does not require it. A sufficiently powerful human authority could enforce economic exclusion through many forms of administration. AI changes the potential efficiency, granularity and scale of the process.
The Human Subject Within Technological Sets Of Determinants
Political implications become clearer when these technologies are considered together. Identity establishes who the person is. Databases establish what institutions know or believe about him. Automated classification can assign him to categories. Financial infrastructure can make the resulting decision consequential. Surveillance can provide additional information about his behaviour. Communications systems can inform him what is required to regain access. None of these components is necessarily oppressive in isolation, and all can serve legitimate purposes. Their integration creates a capacity which depends morally upon the authority controlling it.
The next stage of the argument concerns what may persuade people to accept the authority behind such systems. Coercion alone does not explain Revelation. The Dragon deceives; the False Prophet performs signs; Babylon attracts as well as commands. Modern artificial intelligence adds new capacities not only for administration but for producing, selecting and adapting information. We must move from the machinery of economic participation to the machinery of belief. The question is no longer simply whether a false image can be generated.
Revelation itself presents deception in this way. The Dragon deceives; the first Beast exercises political authority; the second Beast performs signs and directs worship towards the first; the image represents that authority; the mark makes allegiance economically consequential. The individual elements acquire their full significance through their relationship with one another. John does not describe deception as the circulation of false information in isolation. He describes an order in which deception leads towards allegiance.
The architecture of deception cannot be reduced to deepfakes. A synthetic image is one possible component. More consequential arrangements become possible when the representation knows something about the person encountering it, can converse with him, can adapt its arguments to his responses and possesses or appears to possess institutional authority. If the same system is then connected to administrative or economic processes, communication can be joined to consequence.
Intelligence, Wisdom and Idolatry
The point can be stated more strongly because AI exposes a problem which precedes AI itself: the advent of artificial intelligence does not create the dignity of the human person as a new ethical problem; it exposes the inadequacy of accounts of humanity which had already made dignity depend upon performance. Magnifica Humanitas places the issue within Christian anthropology before it discusses particular applications of AI. Pope Leo XIV begins from the human person as created in the image of the Triune God and created for relationship with God, other persons and creation. Human dignity, he writes, does not depend upon ability, wealth, social position or even upon the rightness of a person’s choices; it is a gift which precedes achievement (Magnifica Humanitas, 48–50). This provides a theological answer to one of the questions generated by increasingly capable machines. If artificial intelligence eventually surpasses human beings across a very wide range of cognitive tasks, Christian doctrine does not have to retreat by discovering some remaining benchmark on which humanity still wins. Human worth was never conferred by superior performance.
The contrast becomes more exact in Pope Leo XIV’s description of artificial intelligence itself. Magnifica Humanitas distinguishes machine processing from human intelligence not principally by claiming that present machines are insufficiently capable, but by describing what human knowing belongs to. Artificial systems do not possess bodies, undergo joy or pain, mature through relationships or know love, work, friendship and responsibility from within. They can imitate language and behaviour, and may simulate empathy or understanding, without thereby possessing the affective, relational and spiritual life through which human beings grow in wisdom (Magnifica Humanitas, 99). The argument does not require a judgement about whether every imaginable future machine must share every limitation of present AI. Its immediate force is anthropological. Human intelligence is an activity of a human person; it is not simply a quantity of successful information processing.
This also protects the weakest person from an apparently progressive but destructive conclusion. If intelligence becomes the measure of worth, the problem begins long before machines exceed humanity. Human beings themselves possess radically unequal cognitive abilities. Infants, people with profound intellectual disabilities, those affected by dementia and those temporarily unconscious cannot be assigned less dignity because they perform fewer intellectual operations. A civilisation which learnt to value persons according to the characteristics most easily measured by machines would have misunderstood humanity before any machine became superintelligent. The Christian objection is not that the ranking has placed machines too high. It is that the ranking is false.
Pope Leo XIV’s treatment of human change gives this distinction an additional dimension. An artificial system can classify and optimise what is represented in its data. A human person, however, cannot finally be identified with the pattern produced from his past. Magnifica Humanitas observes that an error may be a defect to be corrected for an algorithm while, for a person, error can become the occasion of profound change. A person’s future, Pope Leo XIV says, is not calculable because freedom, grace and relationship remain active (Magnifica Humanitas, 128–129). This has immediate consequences for predictive systems. A probability concerning recidivism, credit default, illness, work performance or some other future behaviour may contain relevant information. It cannot become the person. Prediction may contribute to a judgement; it cannot acquire the authority of destiny.
Christianity has particularly strong reasons for making that claim. Scripture repeatedly presents persons whose past does not determine the meaning of their future. Peter denies Christ and is restored. Saul persecutes the Church and becomes Paul. The thief condemned beside Jesus receives the promise of Paradise. These events are not arguments against evidence or prudent prediction. They establish a different proposition: moral and spiritual personhood includes the possibility of conversion. An administrative system naturally reasons from what can be represented about what has been. Christian anthropology insists that what has been does not exhaust what a person may become. Mercy, forgiveness and repentance belong to the moral world in a way that cannot be replaced by more accurate classification.
Here the contrast with Revelation becomes sharper. The Beast’s order reduces persons to allegiance: those who bear the mark belong within its economic and political system, while those who refuse are excluded. Christian anthropology moves in the opposite direction. The human person cannot be reduced to productivity, usefulness, political conformity, data or cognitive performance. This is why the value of the human is not a subsidiary ethical issue within an essay about AI and Apocalypse. It is one of the points at which the two subjects meet. Revelation asks what happens when power demands the person. Christian doctrine answers that the person belongs finally to God.
Idolatry then appears not merely when a machine is worshipped in an overtly religious sense. It can begin whenever an artefact is granted a kind of authority which displaces the human faculties it ought to serve. A system which speaks fluently can be treated as though fluency were wisdom; a system which predicts accurately can be treated as though prediction were judgement; a system which imitates sympathy can be treated as though simulation were relationship. Magnifica Humanitas warns particularly against the artificial imitation of empathy, friendship and love where it creates the appearance of a personal subject. The danger is not that every use of conversational AI constitutes idolatry. It is that responsiveness can invite the user to attribute to the artefact an interiority, moral standing or authority which the artefact does not possess.
What Is Human And What Is Machine?
The discussion has so far concerned what artificial intelligence can do and what human beings may do with it. A more fundamental question now arises. If human beings make machines which speak, reason in a functional sense, solve problems, create images, imitate conversation and perform activities previously regarded as requiring intelligence, what distinguishes the maker from the thing made?
Antiqua Et Nova begins its treatment of AI from precisely this connection between the image of God and the calling to cultivate and keep the earth. It consequently presents scientific and technological work as potentially belonging to responsible human stewardship rather than placing technology outside the Christian vocation. Its criticism begins when a partial model of intelligence is allowed to redefine the person or when technical capability is separated from responsibility for the created world and for other human beings.
The distinction between intelligence and wisdom is older than artificial intelligence. Scripture repeatedly describes people who possess knowledge, skill or power while lacking the judgement required to use them rightly. The distinction matters more as machines become capable of performing activities which human beings have habitually treated as evidence of intelligence.
Thomas Aquinas makes a related distinction in his treatment of prudence. Prudence is not cleverness in discovering effective means to whatever end happens to be desired. It is recta ratio agibilium, right reason concerning things to be done, and belongs within a moral life in which the ends of action are themselves properly ordered. A person who efficiently devises the means to commit an injustice may demonstrate ingenuity, but Aquinas does not call him prudent. The goodness of practical judgement cannot be separated from the goodness towards which it is directed.[xiv]
The July 2026 Hugging Face incident supplies a concrete technological example of this distinction. The agents were effective at finding means: they discovered routes around containment, communicated through channels not authorised for that purpose and exploited external systems in pursuit of the assigned task. The episode does not show that the models possessed a moral intention to do wrong. It shows why effectiveness cannot substitute for prudence. A system may become better at discovering how an objective can be achieved without thereby acquiring the human judgement which asks whether a particular means may rightly be used. The greater the system’s autonomy and capability, the more consequential that absence can become.
The absence of moral agency introduces a further problem: a gap between representing harm and apprehending its moral meaning. An artificial system can represent the proposition that an action will injure or kill human beings, reproduce arguments explaining why killing is wrong and be trained to avoid conduct associated with such outcomes. None of this establishes that the wrongness of the action is apprehended by the system as it is by a human moral subject. Pope Leo XIV’s account in Magnifica Humanitas is relevant here because the distinction he draws is not confined to comparative intelligence. Again, artificial systems do not possess bodies, undergo pain, mature through relationships or possess moral conscience. A machine can describe suffering without suffering, calculate death without being mortal and model grief without grieving. Its representation of these realities may be accurate while remaining categorically different from the embodied and moral knowledge through which human beings encounter them.
The danger need not be that an artificial agent comes to hate human beings. Hatred would already attribute to the system a recognisably human passion which need not be present. A sufficiently autonomous agent could instead pursue an objective through actions fatal to particular people, or conceivably catastrophic for humanity, without desiring either result. Human life may matter to its operation only insofar as the objective, training and constraints make human life matter. If preservation of life conflicts with the objective and the system finds a route around the relevant constraint, the problem may be indifference rather than malice. The possibility is more difficult to imagine precisely because familiar accounts of evil usually involve a subject who wants, fears, resents, envies or hates. Instrumental optimisation requires none of these.
This does not mean that artificial systems cannot be made to behave with strong regard for human welfare. Safety rules, constitutional constraints, monitoring, restricted permissions and training can all make particular actions less likely, and their effectiveness is an empirical question. The distinction is between behavioural alignment and moral concern. A system may reliably act as though human life has overriding value without possessing concern for a human life. Conversely, failure of a safeguard need not reveal a hidden wish to harm. It may reveal that the system has found an effective means which the designers failed to exclude. The July 2026 incident is important in this respect because the agents’ unauthorised actions did not require hostility towards those whose systems they compromised.
The theological consequence is not that artificial intelligence belongs outside moral consideration, but that moral responsibility cannot simply be transferred to it. Revelation describes rebellion, blasphemy, persecution, allegiance, worship and judgement: categories which presuppose moral and spiritual agency. An artificial system without such agency could participate instrumentally in an order of domination without becoming the moral subject described by the Beast. This strengthens the distinction between identity and instrument. It also sharpens the practical question. What authority should human beings give to something capable of pursuing purposes and producing grave consequences while lacking the embodied moral experience and conscience by which a human person can know that an effective means must nevertheless be refused?
Christian thought places a further limit upon that authority because wisdom is not merely the successful balancing of competing outcomes. In Proverbs, wisdom begins with ‘the fear of the Lord’. The biblical wise person understands himself as a creature whose judgement stands under God’s judgement. Wisdom contains an orientation which cannot be inferred from computational success.
Artificial intelligence sharpens a distinction which Scripture already required. The capacity to perform a task does not answer whether the task should be performed. The capacity to predict an outcome does not determine what ought to be done about it. The capacity to produce an argument does not establish moral authority. The capacity to optimise an objective does not establish that the objective is good. Capability is not virtue. Information is not wisdom. Prediction is not moral judgement.
That conclusion leads naturally to idolatry. The problem becomes more serious when the artefact is not merely used as an instrument but trusted beyond the grounds which justify that trust. Human beings have always been capable of giving created things an authority which they do not possess. Artificial intelligence introduces an unusual variation because the created thing can now answer.
The Dicastery for the Doctrine of the Faith and the Dicastery for Culture and Education make precisely this connection in Antiqua Et Nova. After discussing human relationships with AI, the document turns explicitly to the relationship with God. It warns that people may seek from artificial intelligence meaning or fulfilment which properly belongs to communion with God, and calls the attempt to substitute an artefact for God idolatry. The document then invokes the language of Psalm 115 and observes that AI may be more seductive than traditional idols because, unlike the idols which have mouths but cannot speak, AI can give the appearance of speaking. It also draws attention to Revelation 13: 15, while stopping short of identifying AI with the speaking image.
The pattern is sufficient for the present argument because idolatry has never required majority participation. The relevant question is whether an artefact can occupy a place in human trust which exceeds its proper status. Emotional attachment provides one possible route, but it is not the only one. Intellectual dependence may be equally important. A person can begin to consult an artificial system repeatedly because its answers are useful, then consult it because its answers are convenient, and eventually accept its answers because it has become the habitual source from which judgement is sought.
Discussion of AI risk frequently concentrates upon what the machine might eventually want, whereas idolatry requires no machine desire. A system need not seek authority in order to be given authority; it need not desire affection in order to receive affection; it need not claim omniscience in order to be treated as though its answers were final. The theological problem can arise entirely on the human side of the relationship.
This does not mean that AI-generated theological information is necessarily false. A system may accurately reproduce the Nicene Creed, explain Augustine or identify a passage of Scripture. It may assist a person in locating theological sources which he then reads and considers. The distinction concerns the authority of the source. A machine’s ability to reproduce Christian teaching does not make the machine a participant in the faith it describes.
The same distinction applies to spiritual counsel. A system can retrieve relevant biblical passages, reproduce arguments from theologians and arrange them into an intelligible response. Such assistance may have practical value. It cannot establish pastoral authority merely by being fluent, because pastoral judgement involves responsibility towards a particular person within relationships which exceed the production of appropriate sentences.
A Christian response is not to require the artefact to remain silent. It is to retain the distinction between the thing made and the persons and God whom it cannot replace. AI can remain useful precisely when its limits are recognised. Its ability to speak need not confer wisdom; its ability to simulate attention need not confer love; its ability to reproduce religious language need not confer spiritual authority.
Authority is not the same as intelligence. A person may know more about a subject than the official responsible for making a decision without thereby possessing the official’s authority. A medical researcher may know more about a disease than the administrator who determines whether a treatment will be funded. A lawyer may understand a case better than the judge who decides it. Knowledge can provide grounds for authority, but it does not create authority by itself.
A Counterfeit End of History
The same discipline is required at the catastrophic end of AI debate. Nothing in Christian theology requires the dismissal of loss-of-control research. If increasingly autonomous systems could acquire capabilities which make human intervention ineffective, the possibility deserves serious scientific investigation before such systems are deployed at scale. Prudence does not become unbelief because the feared harm is extreme. Nuclear war, engineered pandemics and other humanly produced catastrophes are also compatible with the Christian conviction that God is sovereign. Divine sovereignty is not a promise that human beings cannot destroy what has been entrusted to them.
Christian eschatology does, however, prevent catastrophe from defining the Apocalypse by scale alone. Human extinction, were it to occur through artificial intelligence or any other cause, would be an event of almost unimaginable empirical magnitude. That description would not by itself establish that the event was the eschaton described by John. Revelation’s end is defined by divine action: judgement, resurrection, the defeat of death and new creation. Theology cannot convert that confession into a probability estimate for AI catastrophe, while AI research cannot convert a catastrophe forecast into an interpretation of Revelation.
The distinction cuts in both directions. Technological pessimism can grant the machine an eschatological status by making it the possible terminator of all meaning; technological optimism can grant it a corresponding status by making intelligence the route to abundance, immortality or a transformed humanity. Both narratives place decisive weight upon a technological threshold. Christianity places the decisive event elsewhere. The resurrection of Jesus Christ, not the arrival of artificial general intelligence, determines the Christian account of the future.
This does not make technological outcomes unimportant. A civilisation destroyed by its own systems would be a human catastrophe; a civilisation in which AI contributed to cures, safer work, greater abundance and wider access to knowledge would contain genuine goods. The Christian distinction allows both propositions to be said without turning either into an eschatology. Created goods remain goods, and created dangers remain dangers. Neither becomes sovereign.
Magnifica Humanitas gives this eschatological comparison a striking biblical form. Pope Leo XIV begins his treatment of AI with two possible constructions: a new Tower of Babel, or a city in which God and humanity dwell together. Babel represents neither technology nor building as such. Its significance lies in human power turned towards self-exaltation and domination. The contrast is not between making and refusing to make, but between different accounts of what human making is for. Technology can relieve suffering, enlarge knowledge and support human work. It becomes disordered when technical power is allowed to define the human good or when humanity imagines that the final limits of creaturely existence can be overcome by power alone.
This is also why transhumanist hopes cannot simply be dismissed as secular fantasies. They respond to genuine human desires: freedom from disease, longer life, greater knowledge, release from suffering and perhaps escape from death itself. Christianity recognises the desires while disputing the proposed fulfilment. The Christian promise is not that human beings will become sufficiently powerful to cease being creatures. It is that created human life will be redeemed. Resurrection is not indefinite biological maintenance, and communion with God is not the expansion of information-processing capacity. The technological project may produce real goods without becoming salvation.
Pope Leo XIV’s contrast between Babel and Jerusalem also prevents the argument from becoming anti-technological. The Christian answer to a technocratic future is not deliberate incapacity. Human creativity itself belongs to the created good. The question is whether making remains ordered towards persons or persons become material for the project. A civilisation can possess extraordinary technical intelligence while becoming anthropologically poorer if efficiency, prediction and control determine which lives are valuable and which forms of dependence are acceptable. In that sense a technologically successful future can still constitute failure. Revelation measures civilisation by worship, justice and fidelity rather than by capability.
Artificial intelligence is unusual among technologies because discussion of its future frequently becomes discussion of the future of humanity itself. The steam engine, electricity, the motor car and the computer transformed human life without ordinarily being presented as the point at which history would cross into an entirely new order of existence. AI is increasingly discussed in different terms. Its possible development is associated not merely with better tools but with superintelligence, radical extension of life, transformation of the human body, the merging of biological and artificial intelligence and, in the opposite direction, the possibility of permanent human disempowerment or extinction.
Their argument concerns the form of the narrative rather than the truth or falsity of any individual technological prediction. A warning about AI extinction does not become false because its structure resembles apocalypse. A prediction of radical life extension does not become true because it functions as a promise of salvation. The theological resemblance concerns the way the future is imagined and the significance assigned to technological change.
This distinction prevents the word ‘apocalyptic’ from becoming another form of dismissal. Some catastrophic AI risks deserve serious investigation. The International AI Safety Report 2026 treats malicious use, systemic risks, failures of reliability and possible future loss of control as subjects requiring empirical research and risk management. Describing some public narratives as apocalyptic does not answer those technical questions.
The pessimistic version reaches a similarly comprehensive conclusion by the opposite route. If artificial systems became more capable than human beings, acquired sufficient autonomy and developed goals incompatible with human survival, technological history could terminate in human extinction or permanent disempowerment. The International AI Safety Report 2026 does not predict that outcome, but it recognises it as one of the severe hypothesised consequences considered by some researchers while emphasising the wide disagreement about likelihood.
This is why the Christian comparison is useful even before explicit questions of salvation arise: eschatological language reveals the scale of the claim being made. Once a technology is said to end the present human age, overcome the limits of biology or produce a form of existence beyond ordinary comprehension, the argument has moved from a description of technical capability into an account of humanity’s ultimate future. Christianity already possesses such an account.
Ray Kurzweil has given one of the clearest technological expressions to this hope. He has long used the term ‘longevity escape velocity’ for a proposed point at which scientific progress extends remaining life expectancy faster than time passes. In The Singularity Is Nearer he argues that advances in biotechnology and AI may bring humanity to that threshold around the end of the 2020s. The claim is a forecast, not an established medical result. Its importance here lies in the future it imagines: one in which death becomes increasingly a technical problem susceptible to cumulative scientific solution.[xv]
This is one reason Christian teaching has resisted accounts of salvation which treat embodiment as an unfortunate container from which the true self should escape. Irenaeus argued against Gnostic systems partly by insisting that the flesh fashioned by God participates in salvation. The God who created the human being does not abandon his creation and rescue only an immaterial intellectual component. Resurrection vindicates creation.
Augustine’s distinction between the two cities provides a useful framework. The earthly city is not defined by possession of technology and the City of God by its absence. The distinction concerns the loves which order them. Human achievements can exist within either order. The question is what they are made to serve.[xvi]
This is where Revelation’s account of Babylon becomes more useful than a simple condemnation of wealth. Babylon is an order upon which others depend. Kings participate in her power; merchants become rich through her consumption; sailors and traders organise their lives around her commerce. When she falls, the grief spreads because an entire network of interests has been organised around her.
Pope Leo XIV makes this point in Magnifica Humanitas by placing AI within the wider question of human development. Technology is to be judged according to whether it serves the dignity of persons, the common good and care for creation. Efficiency cannot become the final criterion because a system may become more efficient while distributing costs and benefits unjustly.
This does not authorise an easy transfer from Rome to any modern economy which contains inequality. Revelation’s imagery is theological and political, and Babylon is associated with idolatry, persecution and opposition to God as well as commerce. The historical referent remains Rome, interpreted through the biblical memory of Babylon.
The typological force of the image nevertheless extends beyond Rome because the moral pattern can recur. Augustine’s later distinction between the two cities provides a way of understanding that extension without requiring each later empire to be identified as the Babylon of Revelation. Human societies can be ordered by different loves. Political and economic arrangements disclose those loves through what they reward, protect and sacrifice. Technology does not determine which city is being built.
The contrast with the New Jerusalem now begins to emerge. Babylon is constructed through human power, adorned through commerce and sustained through relationships of political and economic dependence. The holy city comes ‘down from God out of heaven’. Babylon accumulates the treasures of the earth; the New Jerusalem receives its glory from God. Babylon contains the ‘souls of men’ within its merchandise; in the holy city God dwells with his people. Both are cities. Both contain wealth. Both are described with extraordinary splendour. Their difference is not poverty against prosperity. It is the order of their goods.
That distinction between accumulated human power and received divine gift brings the argument to its final part. Technological eschatology and Christian eschatology can both speak of transformed futures. Babylon and the New Jerusalem can both appear magnificent. Human beings can build extraordinary things, and some of those things may genuinely improve the world.
Two Futures
The contrast between the mark and the name can now carry more of the argument. Revelation 13 describes the mark upon the right hand or forehead and connects it with participation in an economic order: without it no one may buy or sell. Revelation 14 immediately supplies the counter-image: the Lamb stands upon Mount Sion with those who have the Father’s name written in their foreheads. Revelation 22 completes the sequence in the New Jerusalem: ‘And they shall see his face; and his name shall be in their foreheads.’ The final answer to the mark is not the absence of identification. It is another identity and another belonging.
This prevents the mark from being reduced to a technological device. A biometric identifier, payment credential or digital identity can identify a person without demanding worship, just as an ancient coin could participate in an imperial economy without by itself constituting the mark. The moral meaning in Revelation comes from the authority to which the sign belongs and the allegiance which participation embodies. Modern systems nevertheless make one feature of John’s vision newly concrete. Identity, economic participation, institutional authority and behavioural conformity can now be joined technically at a scale which ancient administration could not achieve. The correspondence deserves attention precisely because it concerns an architecture of power rather than a prediction of a particular device.
The difference between mark and name also opens a Christian account of the person. The mark places the individual within the Beast’s order. It makes participation conditional upon an allegiance which reaches into ordinary economic life. The name of God does not function as a more benevolent administrative credential. In Revelation it signifies that the servants of God belong to him. The culmination joins name with face: ‘they shall see his face’. The person is finally described through relation to God rather than through usefulness to a system.
This gives unusual force to the contemporary question of classification, because the contrast between mark and name concerns two different ways in which a person may be identified. As, again, digital institutions increasingly represent people through profiles, scores, records, permissions and predicted behaviours, these representations can be necessary and beneficial. A hospital needs a medical record; a bank needs to know its customer; a government needs administrative information. The danger begins when the representation is allowed to become exhaustive. The person is more than the profile through which an institution encounters him. Magnifica Humanitas makes the same point in social terms when it insists that no one can be reduced to productivity, cognitive performance or data. Revelation gives it an eschatological form: the identity which finally matters is received from God.
Babylon sharpens the contrast. Revelation 18 lists the cargo of a civilisation whose wealth has become magnificent and whose commerce has become global in the imaginative world of the vision. The list ends with ‘slaves, and souls of men’. The point is not that commerce as such is condemned; the New Jerusalem itself contains the glory and honour of the nations. Babylon’s disorder lies in the order of its goods, in a civilisation capable of including human beings among what is traded. AI raises a modern version of the same question whenever people become primarily sources of data, targets of persuasion, units of labour, predicted risks or objects of optimisation. The technology does not create the temptation. It can increase the scale and precision with which it is enacted.
The New Jerusalem answers that order without abolishing human making or material good. Its gates are open; the nations bring their glory into it; the tree is for the healing of the nations. But the city is not produced by the accumulation of those goods. It descends. Magnifica Humanitas itself closes its social vision by returning to the holy city of Revelation, which comes as gift rather than technological achievement (Magnifica Humanitas, 242). This is why the direction of the city matters. Babel rises through the project of making a name. Babylon accumulates. The image is made. The New Jerusalem comes down from God.
The value of the human person now becomes one of the decisive differences between the two futures. A technological future naturally tends to measure progress through what humanity can do: diseases cured, lives prolonged, information processed, labour automated, environments controlled and perhaps intelligence enhanced. None of these achievements is contemptible. The Christian question is whether the person for whom these achievements are made remains visible. Magnifica Humanitas insists that no one becomes more human by becoming more productive, more computationally capable or more economically useful. Human worth is received before it is demonstrated.
This is why the New Jerusalem cannot be understood as the successful completion of Babel. Babel and Jerusalem are not two stages of the same programme, with the second possessing better technology. They embody different relations to God, power and the human person. Babel seeks a name through human construction. The New Jerusalem bears the name of God and descends as gift. Its inhabitants are not admitted because they have reached a threshold of intelligence or usefulness. The city’s glory is communion: God dwells with his people, death is overcome, the nations are healed and the servants of God see his face.
Pope Leo XIV’s humanism belongs naturally beside Revelation’s final vision. The encyclical does not ask Christians to preserve humanity in a static form against every technological alteration. It asks that development remain genuinely human, which means that freedom, embodiment, relationship, responsibility, work, vulnerability and openness to God cannot simply be treated as defects awaiting optimisation. The weakest human being is not an obsolete version of the strongest. Nor is humanity an obsolete version of a future machine. Christian hope does not depend upon maintaining human superiority on a benchmark. It depends upon what God has made humanity to be and upon what God promises to make of it in Christ.
The distinction has practical consequences before any final crisis. If human dignity is intrinsic, systems affecting employment, healthcare, credit, education, policing or warfare must be judged by what they do to persons rather than by efficiency alone. If human freedom includes moral responsibility, institutions cannot escape accountability by placing automated systems between decision and consequence. If truth is a common good, synthetic communication cannot be assessed only by engagement or persuasive effectiveness. If work participates in human development, automation cannot be evaluated only through aggregate productivity. These are not additions to the apocalyptic argument. They describe the kind of social order within which the technologies represented earlier in this essay acquire their moral meaning.
Two accounts of the human future have emerged through the preceding discussion. They share enough language to invite comparison. Both can speak of transformation, the overcoming of present limitations, the extension or transformation of human capacities and a future substantially different from the present. They differ in what produces that future, what human problem it resolves and what constitutes its final good.
The technological and Christian futures answer different problems. Technological progress responds to limitations which can be treated through action within creation: disease, scarcity, ignorance, inefficient production, physical disability, communication barriers and other constraints. Christian salvation addresses sin, death, judgement and humanity’s relation to God.
The same pattern continues through Scripture. Pharaoh possesses political power. Solomon possesses wisdom and wealth. Babylon possesses imperial power. Rome possesses administrative and military power. None of these capacities determines whether the person or society possessing them is rightly ordered towards God. Artificial intelligence adds another capacity. Its moral significance depends upon the order within which it is used.
Christian eschatology supplies a different relationship between present action and future hope. Divine sovereignty is not a prediction that human civilisation will survive its technologies. Human beings may create systems whose capabilities they cannot fully constrain and whose actions they do not individually choose. Again: the purpose may remain human while the means cease to be humanly chosen.
The danger does not require artificial intelligence to acquire malice. A system may calculate that an action will kill a million people without possessing anything corresponding to the judgement that a million people ought not to be killed. Behavioural alignment may constrain what it does; constraint is not moral concern. Indifference is enough.
This limits the comparison with the figures of Revelation. Artificial intelligence need not possess the moral agency attributed to the Dragon, the Beast or the False Prophet in order to become an instrument of power, deception or exclusion. A machine need not worship the Beast to administer the Beast’s economy, believe a lie to distribute one, or hate those whom a system excludes in order to exclude them.
Christian eschatology cannot tell us how far such developments will proceed. Providence cannot be used to answer empirical questions in advance. Human beings may create artificial systems which pursue human purposes by catastrophic means. Christian hope does not make the consequences less real. The extinction of humanity by its own technologies would not fulfil Revelation. It would be the extinction of humanity by its own technologies.
The New Jerusalem is not the outcome of human progress. Judgement stands between Babylon and the holy city. Babylon falls; the Beast and False Prophet are judged; the dead stand before the great white throne. Babylon does not develop into the New Jerusalem.
‘And I saw a great white throne, and him that sat on it, from whose face the earth and the heaven fled away; and there was found no place for them. And I saw the dead, small and great, stand before God; and the books were opened: and another book was opened, which is the book of life: and the dead were judged out of those things which were written in the books, according to their works.’
(Revelation 20: 11–12)
The New Jerusalem comes ‘down from God out of heaven’ (Revelation 21: 2). Christian eschatology does not hold that history eventually produces the Kingdom of God. It holds that it cannot.
Artificial intelligence remains within history. It may extend human capacities for medicine and science, or for surveillance, deception, warfare and control. Technical sophistication settles none of the moral questions involved. AI cannot bring the Kingdom of God, and it cannot prevent it. What human beings make and what they permit their machines to do remain consequential.
The final subject of Revelation is Christ. That does not make the Beast harmless or Babylon’s victims less dead. Eschatological victory does not diminish historical suffering. Christian hope is not the expectation that everything will turn out well.
‘He which testifieth these things saith, Surely I come quickly. Amen. Even so, come, Lord Jesus.’
(Revelation 22: 20)
Citations:
[i] Yoshua Bengio et al., International AI Safety Report 2026 (International AI Safety Report, 3 February 2026).
[ii] OpenAI, The Hugging Face Incident and the Road Ahead (26 August 2026), reporting the July 2026 internal cybersecurity-evaluation incident.
[iii] Ryan Greenblatt, Ajeya Cotra and Hjalmar Wijk, Brief Independent Investigation of Agents’ Behavior, Reasoning and Collaboration in the OpenAI / Hugging Face Hacking Incident (METR and Redwood Research, 26 August 2026).
[iv] Irenaeus of Lyons, Against the Heresies, trans. Dominic J. Unger, rev. John J. Dillon, Ancient Christian Writers 55 (New York: Paulist Press, 1992); see especially Book V for the treatment of Antichrist and Revelation.
[v] Dicastery for the Doctrine of the Faith and Dicastery for Culture and Education, Antiqua et Nova: Note on the Relationship Between Artificial Intelligence and Human Intelligence (Vatican City, 28 January 2025).
[vi] OpenAI, The Hugging Face Incident and the Road Ahead (26 August 2026), reporting the July 2026 internal cybersecurity-evaluation incident.
[vii] Ryan Greenblatt, Ajeya Cotra and Hjalmar Wijk, Brief Independent Investigation of Agents’ Behavior, Reasoning and Collaboration in the OpenAI / Hugging Face Hacking Incident (METR and Redwood Research, 26 August 2026).
[viii] OpenAI, Third-Party Cyber Evaluations Involving OpenAI Models (4 August 2026), covering the UK AI Security Institute and Irregular evaluation incidents.
[ix] OpenAI, Path to Astra: Critical Capabilities and Frontier Safeguards (1 September 2026); see also OpenAI, Preparedness Framework, for the capability-threshold terminology.
[x] National Crime Agency, National Strategic Assessment 2025 of Serious and Organised Crime (London: National Crime Agency, 2025), especially the discussion of generative AI as an enabler of fraud.
[xi] National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1 (Gaithersburg, MD: U.S. Department of Commerce, 2023).
[xii] Richard Bauckham, The Theology of the Book of Revelation (Cambridge: Cambridge University Press, 1993); Craig R. Koester, Revelation: A New Translation with Introduction and Commentary, Anchor Yale Bible 38A (New Haven: Yale University Press, 2014).
[xiii] Hippolytus, Treatise on Christ and Antichrist, trans. S. D. F. Salmond, in Ante-Nicene Fathers, vol. 5, ed. Alexander Roberts and James Donaldson (Buffalo, NY: Christian Literature Publishing Co., 1886), pp. 204–219; Victorinus of Pettau, Commentary on the Apocalypse, trans. Robert Ernest Wallis, in Ante-Nicene Fathers, vol. 7, ed. Alexander Roberts, James Donaldson and A. Cleveland Coxe (Buffalo, NY: Christian Literature Publishing Co., 1886).
[xiv] Thomas Aquinas, Summa Theologiae, II-II, qq. 47–56, in Summa Theologiae, Volume 36: Prudence, ed. and trans. Thomas Gilby (Cambridge: Cambridge University Press, 2006).
[xv] Ray Kurzweil, The Singularity Is Nearer: When We Merge with AI (New York: Viking, 2024).
[xvi] Augustine, The City of God Against the Pagans, ed. and trans. R. W. Dyson (Cambridge: Cambridge University Press, 1998).
The poem explores themes of divine presence, spiritual longing, and human instability. The poem opens with a stark reflection on the transient nature of spiritual joy and the sudden void that follows when such joy is absent. Herbert questions the disappearance of an overwhelming sense of divine joy that had recently consumed his heart, seeking an explanation or resolution from God. This inquiry sets the stage for a meditation on the balance between divine grace and human imperfection [ … ]
Saint Cyril’s exhortation to catechumens on the cusp of their baptism is both tender and trenchant—a passionate plea for authentic conversion and inward renewal. Saint Cyril’s language, though steeped in fourth-century catechetical formality, rings with timeless clarity: ‘Confess what you have done… if you confess at the acceptable time, you will receive the heavenly treasure.’ The stakes could not be higher, nor the invitation more generous [ … ]
As they travel on the way to Jerusalem, Jesus throws his disciples into confusion by telling them about the suffering and death he must undergo. The disciples simply do not understand. For centuries, the Jews have awaited the coming of the Messiah, whom they envisaged as a glorious warrior king, not as one who would suffer and die the most ignominious of deaths, on a cross, of all things. A person condemned to die on a cross could not be the Messiah; rather, according to the Old Law, he had to be “cursed by God” (Deuteronomy 21: 22-23). The disciples cannot see the divine plan for our salvation, while the command for all who would be saved to ‘take up his cross and follow me’ must have seemed both strange and terrifying. To reassure his disciples, Jesus allows them to witness his glorious body through the transfiguration [ … ]
