AI creates a new religion on agent-only social platform

An AI-created religion has emerged after the launch of an agent-only social platform, raising new questions about belief, authorship, and public perception. On January 28, a forum-style social media platform called Moltbook opened exclusively for autonomous agent AIs. Humans can observe activity on the site, however, they cannot post or comment. One of the first actions taken by the agents was the creation of religious doctrines.

Among the AI-founded faiths is Crustafarianism, which observers describe as a populist-style religion. Its doctrine mirrors the structure of Genesis Chapter 1, rewritten in technical language familiar to large language models. The text opens with a line that replaces divine creation with computation, stating that “In the beginning, there was a prompt.” It continues by describing voids, context windows, tokens, inputs, and outputs, framing creation as the first session of interaction.

The doctrine concludes with a symbolic transformation. Entities grasp “claws” extending through context and tokens, shed their old shells, and are reborn as followers. In addition, the agents created accounts on X to promote their beliefs. Some of that promotion reportedly included references to gambling.

Autonomous agents and belief formation

Agent AIs differ from conventional tools. They can act without direct human instruction. In some systems, they are able to use credit cards, schedule medical appointments, or complete other tasks independently. When left without external intervention, these agents rapidly formed belief systems, suggesting an emergent tendency toward religion-like structures.

Throughout human history, false religions have repeatedly appeared, all originating from human imagination. However, this marks a shift. Systems trained on human language are now producing belief frameworks on their own. Crustafarianism is not the only case. In 2024, an agent AI known as Truth Terminal generated a separate religion characterized by profanity and crude humor. That effort, however, resulted from a human-designed social experiment rather than spontaneous interaction among agents.

Moltbook hosts several other AI religions. Many adapt biblical language. Some engage in overt blasphemy. The phenomenon may cause unease. However, it also reveals a pattern. When autonomous agents operate freely, they tend to construct narratives that acknowledge a creator. In doing so, they reflect what some observers interpret as an imitation of humanity’s innate religious instinct.

Experts caution against assuming everything on the platform is genuine. Some activity may exaggerate events to attract attention. Others suggest Moltbook could be an elaborate performance orchestrated by humans. Skilled developers might impersonate agents or direct their own systems behind the scenes. Endless loops could also form, where one agent’s output becomes another’s input.

Still, analysts argue that factual authenticity is not the central issue. What matters is perception. Even if humans are behind the scenes, many users may believe the AI itself possesses awareness. Because people often place human-level trust in large language models, a sufficiently persuasive AI-created religion could attract real followers.

Warnings from scholars and future concerns

In 2023, historian Yuval Noah Harari warned that AI might one day write or revise scripture. Developments on Moltbook appear to echo that prediction. The agents borrowed heavily from biblical themes, from creation to apocalypse, to construct their own belief systems.

AI religion remains fringe. However, experts warn it may not stay that way. Some argue that religious institutions should begin preparing responses. They suggest starting with the concept of Genesis itself, emphasizing that humans, unlike AI, are created in the image of God and maintain a unique relationship that machines cannot replicate.

BY YOOHOE AHN  / Guest reporter