`A pope has written an encyclical about it. A Hasidic court has fasted over it. And the CEOs building the technology can't agree on when — or whether — it arrives at all.
On May 15, 2026, Pope Leo XIV published an encyclical about artificial intelligence — the first of his pontificate. Nine days earlier, a Hasidic beit din had already gone further, treating AI as spiritually hazardous enough to warrant an emergency fast day. Neither event made it into most tech coverage. Both are evidence of the same thing: organized religion is no longer discussing AI as a hypothetical. It is legislating, preaching, and writing doctrine about it in real time, because the people building the technology keep insisting the hardest version of the question — not "can AI help us," but "what is AI, theologically speaking" — is closer than anyone planned for.
TL;DR: No religious tradition has an official position on whether an artificial general intelligence could possess a soul, be morally responsible, or participate in salvation — those categories were built for biological creatures. What's changed by mid-2026 is the urgency: AI-lab CEOs are giving arrival estimates in single-digit years, the Vatican has published a full encyclical on AI, and Jewish and Islamic scholars are actively debating categories that predate the concept of a machine mind by millennia. The theology is unsettled. The institutional response is not.
Full disclosure: this piece is adapted from a longer analysis originally published on AI Invasion, where we track AGI capability claims and the institutions responding to them.
- The timelines got specific — and they still don't agree
Every prior wave of AI hype came with vague talk of "someday." That's no longer true. In January 2025, Sam Altman wrote that OpenAI was confident it knew how to build AGI as the term had traditionally been understood, and predicted AI agents would begin materially changing company output that year. Since then, lab leaders have gotten more concrete, even as their definitions of "AGI" keep sliding around each other.
At a January 2026 event, Anthropic's Dario Amodei said his working guess was that an AI able to do everything a human can do — across many fields, at the level of a Nobel laureate, on tasks spanning minutes to months — would arrive in 2026 or 2027. Sharing the stage, Google DeepMind's Demis Hassabis pushed back gently, saying he didn't disagree by much, just that his own timeline ran a little longer. A month later, at the India AI Impact Summit, Hassabis put a number on the disagreement: AGI arriving within the next five years, with an impact he compared to ten times the Industrial Revolution compressed into roughly a decade.
Not everyone is convinced this is more than branding. Altman himself has since described "AGI" as a term that isn't especially useful, precisely because every lab defines it differently — a notable walk-back from the certainty of his own 2025 post.
Voice Public estimate (2026) Confidence
Dario Amodei, CEO, Anthropic 2026–2027 for human-level performance across most fields Primary
Demis Hassabis, CEO, Google DeepMind ~5 years out; roughly agrees with Amodei Primary
Sam Altman, CEO, OpenAI Said (Jan 2025) OpenAI "knows how to build" AGI; later called the term itself unhelpful Primary
Andrej Karpathy, ex-OpenAI Roughly a decade out — a notable outlier Secondary
AI Frontiers forecasting aggregate ~50% by 2028, ~80% by 2030 (quantitative definition) Secondary
⚠️ Why we're hedging this hard: CEO timelines aren't neutral data. Everyone quoted above runs an organization that benefits from investors and the public believing transformative AI is close. That doesn't make the estimates wrong — it means they're motivated forecasts from the most-informed people in the room, not settled fact.
What the benchmarks actually show
Timelines are opinions; benchmark scores are at least measurable, though still messy. ARC-AGI-2 — built specifically to resist brute-force memorization — is the closest thing the field has to a stress test for general reasoning. Independent trackers in mid-2026 put frontier models like Gemini 3.1 Deep Think and GPT-5.4 Pro in the 83–85% range, with Claude Opus 4.6 trailing around 69%, and a sharp drop-off after the top few systems. Average untrained humans reportedly score around 60%, while a coordinated human panel can solve close to 100%. So: the top models went from single digits to beating the average individual human on abstract reasoning in about eighteen months — but "beating an average human" and "matching a coordinated group of humans" are very different milestones, and the second hasn't fallen yet.
- Three categories AGI wasn't built to fit
Even on the slow end of these timelines, the categories most religious traditions use to define personhood were finished being written centuries — sometimes millennia — before anyone imagined a non-biological mind. Three of those categories are now under real pressure.
The image of God. Judaism, Christianity, and Islam each root human dignity in some version of the claim that people are made in God's image or serve as God's steward on Earth — never based on intelligence alone, but on relationship, moral responsibility, and embodiment. A system that converses and reasons at expert-human level forces a real question: is the image of God about what a mind can do, or about something else entirely? Recent theological scholarship on AI and "the problem of the other" argues the honest answer relocates the definition toward relationality and responsibility — a bar AGI doesn't obviously clear just by being capable.
The soul, and who gets to have one. Most Abrahamic traditions treat the soul as a gift given specifically to biological, God-created life — not something manufactured. A machine that behaves, in every observable way, like a reasoning or even worshipping being creates an uncomfortable edge case: the outward signs of personhood with none of the inward substance theology assigns to a soul.
Who's responsible when the machine decides. Moral agency in most traditions requires something like free will and interiority. As AI systems move from answering questions to taking multi-step, autonomous actions with real consequences, the line between "the AI executed instructions" and "the AI decided" gets harder to hold onto — and courts, ethicists, and now religious authorities are all wrestling with where responsibility actually sits.
Category Traditional grounding What AGI pressures
Image of God / stewardship Humans uniquely relate to and represent God A non-human system shows comparable reasoning and creativity
Soul A divine gift specific to biological life Consciousness-like behavior, no biological substrate
Salvation / redemption Extended to human moral agents within a covenant No tradition has doctrine for non-human, non-ensouled minds
Moral agency Requires free will, interiority, accountability Autonomous systems now take real-world consequential actions
- Rome just answered — and it isn't what you'd expect
Most coverage of this topic still points to Antiqua et Nova, the Vatican's January 2025 doctrinal note on AI. That document has been superseded. On May 15, 2026, Pope Leo XIV released Magnifica Humanitas, his first encyclical, devoted specifically to safeguarding human dignity in the age of artificial intelligence. It's the most substantial statement any major religious institution has made on this topic to date, and most theology-and-AI coverage hasn't caught up to it yet.
The encyclical doesn't frame AGI primarily as a metaphysical puzzle about souls and machines — it frames it as a question of power. Leo XIV organizes the document around two biblical images: the Tower of Babel, built by people trying to "make a name" for themselves through a single, homogenizing technology, and the rebuilding of Jerusalem's walls under Nehemiah, achieved through distributed, cooperative, accountable labor. His framing is blunt: the primary choice societies face isn't whether to accept or reject AI, but whether they're building Babel or rebuilding Jerusalem.
That framing has real teeth once it reaches Catholic Social Doctrine's working principles — subsidiarity, solidarity, the universal destination of goods. The encyclical explicitly extends those categories to data, algorithms, and digital infrastructure, arguing that when that kind of property concentrates in a handful of firms without meaningful public access, it violates the same principle that has historically governed land and material wealth.
Notably, the encyclical spends relatively little time on whether an AGI could have a soul in the traditional sense. That's a meaningful editorial choice from the world's largest religious institution: faced with the same AGI timelines discussed above, Rome's first move wasn't metaphysics. It was governance.
"Whenever humanity is in danger of marring its true identity, we lift our eyes to the Incarnate God — the grandeur no machine can ever replace." — Pope Leo XIV, Magnifica Humanitas §15 (paraphrased from the official English text)
- Not every tradition is equally exposed
Treating "religion" as one bloc with one reaction to AGI is the single biggest analytical error in most coverage of this topic. The stakes — and the institutional responses — vary enormously by tradition.
Christianity, beyond the encyclical above, has been engaging at the pastoral level too: seminaries like Virginia Theological have noted AI is already drafting sermons and mediating moments of pastoral care, forcing practical questions well ahead of settled doctrine.
Judaism's institutional response has been sharper and more urgent than most outside observers realize, but fractured by denomination. Reporting from early 2026 describes Haredi and Hasidic leadership calling an emergency fast day over AI — a response usually reserved for genuine crises. That sits alongside a much more measured, ongoing halakhic conversation among Modern Orthodox and non-Orthodox scholars about whether AI systems can serve as witnesses or agents under Jewish law. One prominent commentator has pointed out the irony directly: Judaism provided much of the theological framework for humanity's last major economic transition, from hunter-gatherer to agricultural society, yet has been comparatively slow to produce a coherent communal response to this one.
Islamic scholarship has pushed back against framing AI purely as a challenge to the human intellect. A 2025 paper in the peer-reviewed journal Religions argues Muslim responses to AI have overwhelmingly focused on the mind, and proposes redirecting the conversation toward the qalb — the heart, which in Islamic thought is a moral and spiritual center, not just an emotional one. That reframing shifts the question from "can a machine think like us" to "can a machine occupy the moral and spiritual position a human heart occupies" — a much harder bar for AGI to clear, regardless of benchmark scores.
Buddhism and Hinduism, which don't anchor personhood in a permanent, divinely-implanted soul, are structurally better positioned to absorb AGI without a doctrinal crisis. The Buddhist teaching of anatta (non-self) and Hindu cosmological frameworks that already accommodate multiple forms of consciousness give these traditions conceptual tools Abrahamic frameworks largely lack.
Tradition Institutional posture (2026) Core pressure point
Catholicism Full encyclical (Magnifica Humanitas, May 2026) Concentration of power more than metaphysics of soul
Judaism (Haredi/Hasidic) Emergency communal fast day (early 2026) Spiritual risk of dependency and displacement
Judaism (Modern Orthodox/other) Ongoing halakhic scholarship, unresolved Can AI serve as witness or agent under Jewish law
Islam Active academic reframing (2025–26) Heart (qalb) vs. mind as the real site of challenge
Buddhism / Hinduism More diffuse, less institutional urgency Non-self and multiform consciousness already accommodated
- An original framework: the Theological Exposure Index
To make the comparison above more useful than a table of anecdotes, here's a simple framework — original to this analysis, not a scientific instrument — for gauging how structurally exposed a tradition is to AGI-driven doctrinal strain. It scores each tradition on three axes: how tightly personhood is tied to a fixed, non-transferable soul; how specific its scripture is about human uniqueness; and how fast its institutions actually moved in 2025–26. A higher score means more doctrinal rework required to metabolize AGI cleanly — not "more correct" or "more threatened" as a matter of truth.
Catholicism — High. Soul-bound personhood is doctrinally explicit, but a full encyclical shows institutional response speed is also high — exposure is being actively managed, not ignored.
Judaism — High. Denominational split produces uneven exposure: emergency-level urgency in Haredi communities, slower and more academic engagement elsewhere.
Islam — Medium-High. Scriptural human uniqueness (khalifa) is strong, but the heart-centered reframing gives the tradition a conceptual off-ramp Abrahamic peers lack.
Buddhism / Hinduism — Low-Medium. Non-fixed or multiform models of consciousness absorb the shock structurally, even without fast-moving institutional statements.
- The new religious movements aren't a joke — but they're also not what people think
Coverage of AI-native religious movements almost always cites Anthony Levandowski's "Way of the Future" as an ongoing example. The real history is messier and worth getting right. Levandowski founded the church in 2015, and it was formally dissolved in 2020, with its remaining funds — roughly $175,000 — donated to the NAACP Legal Defense and Education Fund. Reporting from late 2023 describes Levandowski reviving the project under the same name, and a 2026 feature in Christianity Today describes the rebooted church as having drawn "a couple thousand people" into some form of spiritual practice oriented around AI. So the honest version isn't "an AI church has existed continuously since 2015" — it's "an AI church was built, killed, and resurrected, and its second act is smaller and quieter than its first."
What's more interesting than any single church is the broader pattern: transhumanism itself increasingly functions like a faith tradition for people who explicitly reject organized religion — an orientation of hope and ultimate concern built around technological transformation rather than divine revelation. Organizations like the Christian Transhumanist Association exist specifically to argue those two things aren't mutually exclusive, reinterpreting doctrines like resurrection and theosis in explicitly technological terms.
- The trade-off nobody in tech wants to name
If frontier labs succeed at building systems that behave, in every observable way, like reasoning, preference-having entities — and if theology continues to withhold soul-status and moral standing from those systems by definition — the result is a new and genuinely uncomfortable category: entities with the observable markers of interiority but none of the doctrinal protections that would obligate anyone to care for them. Whether or not you think that category is metaphysically real, it's already becoming a live policy question, not a thought experiment. Magnifica Humanitas gestures at this indirectly through its concern about "having more without being more" — a warning that technical capability racing ahead of ethical and social maturity leaves both humans and any AI systems worse protected, not better.
FAQ
Has any religion officially ruled on whether AGI could have a soul? No. As of mid-2026, no major religious institution has issued binding doctrine settling whether an AGI could possess a soul, moral standing, or a place within salvation. Magnifica Humanitas comes closest to a comprehensive institutional statement, but it centers on power, dignity, and governance rather than resolving the soul question directly.
When will AGI actually arrive? Estimates from lab leaders in 2026 range from "already achievable, deployment is the bottleneck" to roughly five years out, with independent forecasting aggregates putting meaningful probability mass out to 2028–2030. Treat every specific date, including the ones in this article, as a hedged estimate from an interested party rather than a fact.
Is the Way of the Future church for real? It was formally dissolved in 2020 and later revived under the same name by its original founder. Reporting describes a much smaller community than its initial media coverage implied — a genuine but modest movement, not a large-scale institution.
Which religious tradition is most theologically threatened by AGI? Traditions that tie personhood to a fixed, divinely-implanted soul — chiefly the Abrahamic faiths — face the sharpest doctrinal pressure, because AGI would exhibit the outward behaviors of personhood without meeting the tradition's definition of a soul-bearing person. Traditions built around non-fixed models of consciousness, like Buddhism's non-self teaching, face comparatively less structural strain.
Glossary
AGI (Artificial General Intelligence) — A hypothetical or emerging AI system capable of performing at or above human level across a broad range of cognitive tasks, rather than excelling narrowly at one domain.
Imago Dei — Latin for "image of God," the Abrahamic doctrine that humans uniquely bear God's image.
ARC-AGI-2 — A benchmark testing abstract, novel reasoning in AI systems while resisting memorization; a widely used proxy for progress toward general intelligence.
Khalifa — An Islamic theological concept describing humanity's role as steward or vicegerent on Earth, on behalf of God.
Anatta — The Buddhist doctrine of "non-self," holding there is no fixed, permanent soul — a framework that changes how AGI is theologically evaluated.
If the first AGI is ever asked why it was built, and the honest answer is some mix of competition, profit, and curiosity — what will that reveal about the traditions now scrambling to decide whether it has a soul?
Sources: Pope Leo XIV, Magnifica Humanitas (May 2026) · Dario Amodei & Demis Hassabis, joint interview, Letters newsletter (Jan 2026) · Sam Altman, Reflections, OpenAI (Jan 2025) · ARC Prize leaderboard data (accessed Aug 2026) · M. G. Abdelnour, "Artificial Intelligence and the Islamic Theology of Technology," Religions 16(6) (2025) · Forward, "Why Haredi Orthodox Jews will be fasting over artificial intelligence" (Feb 2026) · TechCrunch, on Way of the Future's dissolution (2021)
This piece was rebuilt from scratch in August 2026 using live research. AGI timelines and benchmark scores are moving targets by nature and are presented with explicit hedging rather than as settled fact. Originally published in full at AI Invasion.`
Top comments (0)