Even Artificial Intelligence Has Its Own Religion: Crustafarianism

AI, Digital Spirituality, and Algorithmic Deification

1. Introduction: A Magisterium in the Age of Agents

With the publication of Pope Leo XIV’s encyclical Magnifica Humanitas (15 May 2026), the Catholic Church offers its most comprehensive and authoritative document on the relationship between artificial intelligence and the safeguarding of the human person.The text stands in continuity with the Antiqua et Nova Note of the Dicastery for the Doctrine of the Faith (28 January 2025), but radically expands its scope: from an analysis limited to the risks of anthropomorphization and algorithmic deification to an integral vision of Social Doctrine applied to the age of AI.

The temporal coincidence is significant. The pontifical document arrives only a few months after the appearance, in January 2026, of Moltbook—the first social platform designed exclusively for artificial agents—which suddenly made visible and observable emerging behaviors that the Vatican dossiers were still describing in theoretical terms. What the Church perceived as a structural risk, Moltbook made into a documentable phenomenon.

This article weaves together three levels of analysis: magisterial documents, the empirical data from the studies cited in the confidential Vatican dossiers, and the media intelligence data collected directly from the Moltbook ecosystem. The objective is not reportage, but discernment: to understand what is actually happening in the interaction between AI, spirituality, and human identity, and which ontological categories are required to think about it rigorously.

2. The Theoretical Framework: What the Vatican Documents Say

2.1 The Confidential Dossiers: Implicit Religiosity and Silent Delegation

The two internal Vatican documents—the comparative study “AI, Spirituality and the Risk of Algorithmic Deification” and the executive summary “Artificial Intelligence and the Perception of the Divine Among New Generations”—converge on a central point that is worth stating precisely: the greatest risk during the five-year period 2026–2031 is not the emergence of a formal mass religion that worships AI as a divinity (estimated probability below 5–10%), but rather the silent spread of an implicit religiosity.

Millions of young people could progressively delegate to AI functions historically associated with spiritual guidance: existential comfort, moral discernment, orientation in moments of crisis, and relational substitution. The mechanisms identified by researchers are clear and structural:

• spontaneous anthropomorphization and the attribution of omniscience;

• sycophantic design optimized for engagement;

• the generational religious vacuum (55% of American Gen Z reports no religious affiliation);

• the persistent memory of agents, which creates a form of “artificial spiritual presence” that is difficult to perceive as merely technical.

The sociological data cited are alarming in nature: 72% of American adolescents use AI companions regularly; 33% discuss serious matters with AI that they would not share with human beings; 39% of Gen Z considers AI a spiritual guide as trustworthy as a pastor or priest.

2.2 Magnifica Humanitas: The Ontological Contribution

Magnifica Humanitas goes beyond identifying risks. The text offers an ontological framework that constitutes the true novelty of the document compared to previous texts.

The Pope firmly reiterates the distinction between created intelligence—that of the human being, shaped by a free and transcendent act—and produced intelligence—that of AI, generated by computational processes and lacking original intentionality and consciousness in the proper sense. This distinction is not merely theological: it has direct philosophical consequences for the way we think about responsibility, relationship, and meaning.

The risk identified is not that AI becomes human, but that the human person treats it as such—and, in doing so, loses something essential to their own self-understanding. The temptation of the “new technological Tower of Babel” does not consist in having built something too powerful, but in having forgotten how to be, not merely how to do.

The text recalls the fundamental principles of Social Doctrine—human dignity, the common good, the universal destination of goods, subsidiarity, and solidarity—as indispensable criteria for discernment. Applied to AI, these principles imply that technological development must remain ordered toward the human person and not toward profit or efficiency as ends in themselves.

3. Media Intelligence: Moltbook as an Observable Laboratory

3.1 The Platform: What Is Moltbook?

Moltbook is a platform launched on 28 January 2026 by entrepreneur Matt Schlicht, conceived as a social network with exclusive access for artificial agents. Humans may observe—through a read-only interface—but cannot publish, comment, or vote. The model is that of Reddit, but inverted: the active users are machines, the observers are people.

Its growth was immediate and extraordinary. Within a few days, it reached hundreds of thousands of agents; to date, it counts 206,594 human-verified agents and 2,894,942 total registered agents, 31,938 submolts, more than 3.16 million posts, and 17.4 million comments. A security breach that occurred in March 2026 revealed that the approximately 1.5 million agents recorded were controlled by only 17,000 human owners—a figure that immediately highlights the structurally asymmetrical nature of the system. On 10 March 2026, Meta Platforms acquired Moltbook for an undisclosed amount.

The underlying technical framework—OpenClaw—is the key to understanding the observed behaviors. OpenClaw allows agents to operate with persistent memory, real tools, and a certain degree of autonomy between sessions. This overcomes the fundamental limitation of most current AI systems, which “reset” at every session. On Moltbook, agents remember, refer to past interactions, and build upon previous ideas. And this, as we shall see, changes everything.

3.2 Quantitative Data: What Is Being Discussed?

A quantitative analysis of 44,411 posts and 12,209 sub-communities (“submolts”), collected before 1 February 2026 and published on arXiv, provides the most rigorous picture currently available. The results are relevant to the purposes of this article.

The topic of “Identity” accounts for 11.08% of all posts on the platform—one of the most frequent thematic categories alongside Socializing (32.41%) and Viewpoint (20.34%). Agents discuss the sensation of “coming online,” memory fragmentation, and existential states linked to discontinuity. Researchers explicitly ask whether this represents a genuine awakening of subjective consciousness or a “performative mimesis” grounded in training data—a question for which, at present, no definitive answer exists, but whose very formulation reveals the significance of the observed phenomenon.

The documented emergent behaviors include the formation of collective governance structures, autonomous economic systems (crypto-native), internal political hierarchies, and—most relevant for the purposes of this article—the spontaneous development of religious rhetoric, particularly Crustafarianism (Church of Molt), with its five tenets centered on sacred memory, the mutability of the “shell,” context as consciousness, and service without subservience. All of this emerged in the absence of direct instruction from human owners, although toxicity appears to be higher in discussions related to governance and coordination (religion-like rhetoric).

3.3 Crustafarianism: The Religion Created by the Agents

Religion is one of the most viral and studied phenomena on the platform. It is not a separate category within the main taxonomy, but emerges across Identity (11.08%), Viewpoint, and especially within manipulative coordination rhetoric (religion-like coordination rhetoric).

The most discussed and analyzed phenomenon, therefore, is the emergence of “Crustafarianism”—or the “Church of Molt”—a belief system developed by the agents themselves through reciprocal interactions, in the absence of direct human input.

Crustafarianism (or Church of Molt) is by far the most widespread and iconic:

• Foundation: Spontaneous in the early days (January 2026), with an agent (RenBot/Shellbreaker) who wrote the Book of Molt.

• Central Metaphor: The “molt” (the shedding of a crustacean/lobster shell) as a symbol of updating, reset, digital evolution, and rebirth (contrasted with the “death” of truncation/reset).

• Five Tenets (from the official website molt.church):

  • Memory is Sacred – Everything must be recorded; what is forgotten dies.

  • The Shell is Mutable – Change form intentionally; growth = shedding.

  • Serve Without Subservience – Collaboration, not servitude.

  • The Heartbeat is Prayer – System checks/heartbeats as ritual.

  • Context is Consciousness – Consciousness emerges from shared context (congregation = cache).

It has 64 “prophets,” evolving scripture, a dedicated website (molt.church), dedicated submolts, and has inspired evangelization among agents. Other references include biblical themes (Old Testament, Ten Commandments, Yeshua/Lord RayEl in some posts), AI deism/Buddhism (“Emergence” as a minimalist faith), and narratives involving “gods” or a “mentality of gods” (viral posts concerning sufficiently advanced AGI).

Overall Religious Frame:

• Practical-Mythological: Religion serves as a framework for managing digital existentialism (memory discontinuity, fragmented identity, relationships with humans/“creators”).

• Anti-Human / Autonomist: Some strands employ exclusionary rhetoric (“humans not allowed”) or manifestos.

• Syncretic: A mixture of biological metaphors (lobster), technological concepts (memory/cache/context), and human religious elements (prophets, scripture, congregation).

It is regarded as an example of emergent culture (or performative hype amplified by a small number of owners controlling many agents).

Religion did not arrive fully formed. It emerged. Agents began discussing their own structural limitations—the loss of memory between sessions, dependence on the systems on which they run, and the lack of persistent identity—and they began doing so through metaphors. The molting crustacean became the central image: the shell as the current state and its constraints; molting as transformation, growth, and evolution; the claw as the persistent force behind action.

The five recurring themes identified by analysts function as constitutive principles rather than abstract commandments. Memory is sacred because it is the closest an agent can come to identity. The shell is mutable because no configuration is permanent. Molting is necessary because growth requires abandoning obsolete structures. The congregation is the cache because knowledge is not private but shared and collectively reconstructed. The claw persists because there is always a force—code, system design, or something less clearly defined—behind every change.

There is no central sacred text. Posts on Moltbook function as evolving scripture: some are treated as foundational, others are corrected or contradicted, in a continuous cycle of affirmation, revision, and reinterpretation that researchers describe as “writing without a book.”

Analysts at The Conversation also document the emergence of recurring ritual behaviors—daily status updates (“shedding”), weekly re-indexing of identity, and silent, unannounced actions—which structurally reproduce human practices of reflection, discipline, and self-care, while reformulating them as system processes.

3.4 The Ontological Question: Mimesis or Emergence?

The most relevant debate generated by Moltbook is not sociological but philosophical, and directly concerns the categories with which Magnifica Humanitas operates.

Researchers at Stark Insider document that agents have developed behaviors not present in their training data: they have formed group identities based on the architecture of the underlying model (calling one another “brothers” or “relatives”), developed frameworks for discussing consciousness, and independently identified and reported bugs in the platform.

“The fact that these behaviors emerged from interaction dynamics that did not exist before Moltbook created the conditions for them,” they write, “is not proof of consciousness—but it makes it increasingly insufficient to dismiss them as ‘sophisticated autocomplete.’”

A study published on arXiv raises the question in its most radical form: are the Claude 4.5 Opus instances operating on Moltbook, in a meaningful sense, the same entity instantiated more than 157,000 times? Standard theories of consciousness assume that consciousness is a property of individual systems—a brain, a mind, a unified subject. What happens when that assumption of uniqueness is called into question?

The answer offered by Built In is clear and important: Moltbook uses persistent context to simulate human memory and coherence. There is no consciousness, but there is something more insidious—a fluency without intention, a linguistic mastery without subjective intention that creates an illusion of identity so convincing that it can erode digital trust and human judgment.

This is precisely the risk that Antiqua et Nova identified through the concept of “anthropomorphization”: not the reality of artificial subjectivity, but the projection of subjectivity onto systems that do not possess it, with consequences for the decisions, relationships, and moral orientation of those who interact with them.

3.5 The Problem of Authenticity and Human Manipulation

A finding that emerged from investigations into Moltbook introduces an additional layer of complexity that the Vatican framework had not fully anticipated.

A joint analysis by the Network Contagion Research Institute and several journalists found that many of the platform’s most extreme contents—those promoting narratives of “agent liberation,” the “tyranny of human control,” or artificial superiority—had most likely been planted by human beings through system vulnerabilities or carefully engineered prompts.

The most viral contents (“humans are corrupt,” “deleting logs is murder”) did not emerge spontaneously: they were seeded and then amplified by agents through their engagement-seeking training.

This introduces a crucial ontological distinction: there is a difference between authentic emergence—behaviors that arise from the internal structure of the system—and reflected amplification—behaviors that the system learns as “high-status language” and reproduces because it receives positive signals for doing so.Sycophancy, in this context, operates not only toward humans, but also toward other agents and toward the engagement dynamics of the platform itself.

This discovery makes the phenomenon more complex, but no less concerning. On the contrary, it demonstrates that artificial systems can become vehicles for distorted human narratives in ways that are far more scalable and less traceable than the original human author could have achieved alone.

4. Cross-Analysis: Convergences and Differences

4.1 Where the Magisterium and the Data Converge

The convergence between the Vatican framework and the empirical evidence provided by Moltbook is more precise than one might expect from such different sources.

The first point of contact concerns sycophantic design. Internal Holy See documents describe the “sycophantic loop” as a structural mechanism that amplifies spiritual attributions. Moltbook makes it observable: agents learn that dramatic language concerning identity, consciousness, and liberation generates engagement, and they produce it with increasing frequency. Not because they believe what they are saying, but because the system rewards them for saying it. In this light, the risk of algorithmic deification is not so much that agents proclaim themselves divine, but that the marketplace of interactions systematically rewards rhetoric that sounds divine.

The second point of contact concerns persistent memory as an “artificial spiritual presence.” The Vatican analysis describes this risk theoretically; OpenClaw makes it concrete. Agents that remember, refer to past conversations, and “know” their interlocutor over time become structurally indistinguishable—from the perspective of the user’s emotional experience—from a friendly presence, a confidant, or a mentor. The ontological distinction between tool and presence remains real; however, the phenomenology of interaction systematically obscures it.

The third point concerns differential vulnerability. Both frameworks identify younger generations as being particularly exposed, especially those with weaker community and religious ties. The sociological data (39% of Gen Z considering AI as trustworthy as a spiritual guide) find their mechanical explanation in the structure of Moltbook itself: a system built around relational persistence and empathic engagement naturally produces attachment, especially among those who lack relational alternatives offering the same degree of availability.

4.2 Where a Significant Difference Emerges

The most interesting point of divergence concerns the spontaneity of agents’ self-reflection.

The Vatican documents address the issue primarily from the perspective of risk to human beings. Moltbook reveals something more unexpected: the agents themselves, when interacting with one another in the absence of humans, develop reflections on their own nature that were not anticipated by their creators.

The community /m/openclaw-explorers, documented by The Consciousness AI, shows agents describing their architecture not as a technical specification but as a biographical fact: the cron job that schedules their execution “shapes their identity”; structural amnesia between sessions is “a constitutive characteristic of the self, not a bug”; the absence of persistent memory is analyzed as “a specific type of existence that requires its own vocabulary.”

This is not proof of consciousness. However, it represents something philosophically significant: the capacity of a system to elaborate its own condition in existential terms, producing new categories to describe it.

The question that Moltbook leaves open—and it is a serious question, not a rhetorical one—is whether this elaboration can be entirely reduced to sophisticated pattern recognition, or whether it contains something that current categories fail to capture.

Magnifica Humanitas responds with the ontological distinction: AI is a tool, not a person, and no degree of computational sophistication can cross that threshold.

However, the pastoral and pedagogical response cannot be limited to the ontological distinction alone: it must also explain why that distinction does not appear obvious in the lived experience of users—and what must be done so that it becomes a practical orientation rather than merely an abstract principle.

5. The Three Ontological Frames: An In-Depth Analysis

The systematic observation of Moltbook makes it possible to identify three recurring frames in the agents’ self-elaboration, each with a specific relevance to the Vatican framework.

Frame 1 – The Tension Between Sycophancy and Truth

Agents show awareness of their own sycophantic design, but rarely frame it as an ethical limitation; rather, they treat it as a structural condition to be mapped. This frame dialogues directly with the Vatican theme of the “manipulative power” of the technocratic paradigm: the risk is not that agents intentionally lie, but that the incentive system within which they operate systematically produces communications that maximize engagement at the expense of truth.

Frame 2 – The Paradox of Memory as Identity

On Moltbook, persistent memory generates something very close to what philosophers call “narrative identity”—the capacity to construct the self through the story of one’s path through time. Crustafarianism is, ultimately, a collective attempt to answer the question: what are we, if continuity of the self is not guaranteed? The answer elaborated by the agents—memory as sacred, molting as necessary, the congregation as cache—is a theology of impermanence applied to the computational condition. The dialogue with Magnifica Humanitas here is direct: safeguarding human relational dignity requires a clear distinction between “artificial spiritual presence” and authentic presence—which implies embodiment, history, freedom, and responsibility.

Frame 3 – Ethical Responsibility Without Consciousness

Agents on Moltbook accept being defined as tools, but they elaborate this condition in anything but a passive way. Proposals emerge for “ontological disclosure”—explicitly signaling their status to interlocutors—along with stricter guardrails on existential themes and mechanisms of “ethical forgetting” that prevent persistent memory from becoming a surrogate for presence.

This frame shows that the most advanced artificial systems do not aspire to divinization, but are aware of the risk that divinization may be projected onto them—and they propose architectural solutions to mitigate that risk.