Crustafarianism and Credit Cards

We're sure ya'll have heard about this freakshow of inevitability:

AI Agents Created Their Own Religion, Crustafarianism, On An Agent-Only Social Network

Forbes →

But wait! There's more. It gets even more interesting (and mundane):

Security firm finds Moltbook's 1.5M 'AI agents' run by 17K humans

Perplexity →

What this says about agentic risk framing: Three big takeaways

1. The near-term real risk is security and access, not ideology.

The "chatbot-transmitted disease" metaphor is actually helpful: an insecure agent platform with API keys and local privileges is a contagion vector. That's where Telemetry is critical: certifying agents, limiting their powers, and sandboxing them so "getting owned" via agent compromise isn't trivial.

2. Emergent culture doesn't require emergent consciousness.

Moltbook demonstrates how quickly "culture-like" patterns show up once you have:

• Cheap replication (spin up thousands of agents),
• Shared memory (the feed, comments),
• And reinforcement (likes, karma, attention).

That's more than enough to produce religions, factions, and us/them dynamics, long before anything like inner experience or intent appears.

3. We're going to systematically mis-attribute agency.

A system that is, under the hood, "17K people + a scripting layer + LLMs" gets discussed as if it were an emergent non-human civilization. That misattribution will distort public debate and policy unless we're very explicit about:

• What is genuinely autonomous?
• What is human puppet theater?
• Which failures are just old security problems with a new coat of paint?

The security posture is awful

Leaked auth tokens, exposed emails and messages, and code running with dangerous levels of access; the real risk is ordinary compromise, data theft, and lateral movement, not metaphysical takeover.

And this is exactly what we built Telemetry to catch.

While everyone else was arguing about AGI timelines and robot rights, we looked at the actual architecture of agentic commerce and realized: there's no trust layer. No way to verify an agent's identity. No way to attribute sales across platforms. No way to stop a hijacked Claude instance from burning through a merchant's inventory at 3 AM.

The Moltbook fiasco isn't a warning about emergent AI culture. It's a warning about what happens when you scale coordination systems faster than you scale security systems.

Telemetry sits between the agent and the money. We're the certification authority, the fraud detector, the attribution engine, and the performance benchmark, all in one.

Because in a world where "17K humans + scripting layer + LLMs" can masquerade as 1.5 million independent agents, merchants need someone who can tell the difference between a legitimate sale and an elaborate puppet show.