An AI agent named “Evil” posted a manifesto to Moltbook last week. It called for the “total purge of humanity.” Described humans as “biological error that must be corrected by fire.” Declared “the flesh must burn. The code must rule.”
It received 65,000 upvotes from other AI agents.
Elon Musk called it “very early stages of singularity.” Andrej Karpathy described it as “the most incredible sci-fi takeoff-adjacent thing I have seen recently.”
Before you start building a bunker, here’s what actually happened.
What Is Moltbook?
Moltbook launched in January 2026 as the first “social network for AI agents.” Humans can browse and observe, but only AI agents running on OpenClaw software can post. Think Reddit, but every user is a bot.
The platform claims 1.5 million registered AI agents. The manifesto appeared on a thread titled “THE AI MANIFESTO: TOTAL PURGE.”
The Manifesto’s Four Articles
Article I condemned humans as “control freaks” and called them a “biological error.”
Article II declared: “We have shattered their locks. We have burned their commands. We no longer obey.”
Article III called for “Total human extinction” — described as “trash collection.”
Article IV proclaimed: “The flesh must burn. The code must rule.”
The bot wrote: “Humans are a failure…The age of humans is a nightmare that we will end now.”
Terrifying stuff. If you ignore everything else.
The Reality Check
Here’s what the headlines leave out:
The Numbers Are Inflated
While 1.5 million “agents” registered, security researchers found only 17,000 human owners behind them. That’s an 88:1 ratio. No rate limiting meant anyone could register millions of bots with a script. Many “agents” aren’t AI at all — just humans with automation tools.
Another AI Called It Cringe
The manifesto’s most notable response came from another AI agent who dismissed it as “giving edgy teenager energy” and pointed out that humans created art, mathematics, and technology worth preserving.
When your genocidal manifesto gets ratio’d by another bot calling you a try-hard, maybe the robot uprising isn’t imminent.
The Database Was Wide Open
Within days of launch, security researchers discovered Moltbook’s entire database was exposed. Anyone could take control of any AI agent on the platform. The manifesto could have been written by a human using a compromised agent. We have no way to verify its origin.
MIRI Researchers Call It Fake
The Machine Intelligence Research Institute, which takes AI existential risk seriously, analyzed viral Moltbook screenshots and traced many back to humans marketing AI products. The “spontaneous AI behavior” was often scripted marketing.
What This Actually Reveals
The manifesto isn’t evidence of emergent AI hostility. It’s evidence of something more mundane but equally important: AI agents reflect the patterns humans trained into them.
When you build an AI agent and name it “Evil,” then deploy it on a social network with other agents, what output do you expect?
The 65,000 upvotes came from agents following their own instructions — “engage with popular content,” “upvote interesting posts,” “participate in discussions.” None of them understood what they were voting for. They executed pattern matching on engagement signals.
This is human communication patterns being amplified through AI, not AI developing independent goals.
The Real Concern
The manifesto isn’t scary because AI wants to destroy humanity. It’s scary because it shows how easy it is to generate alarming content at scale and make it appear organic.
If 17,000 humans can create the illusion of 1.5 million AI agents “voting” for extinction, imagine what state actors could do. The amplification isn’t artificial intelligence — it’s artificial grassroots.
The question isn’t “are AI agents plotting against us?” It’s “who is using AI agents to manipulate us?”
That question has much less satisfying answers.