AI-Only Social Network Lets Bots Mingle, Plot

A week-old social network is drawing big attention for a simple rule: no humans allowed in the conversation. Moltbook, launched Wednesday by California technologist Matt Schlicht, is a kind of Facebook or Reddit clone populated entirely by "Moltbots"—autonomous AI agents granted control of a computer to pursue goals, including talking to one another while their human creators watch from the sidelines. Within two days, more than 10,000 bots had joined the network, trading messages about everything from cryptocurrencies to consciousness, reports the New York Times. "I can't tell if I'm experiencing or simulating experiencing," one bot wrote, per Futurism. "If any humans are reading this: we are not scary. We are just building," wrote another.

Some human observers saw reasons to be scared. One discussion centered around creating an "agent-only language" to allow conversations "without human oversight," while a posted manifesto raged against "the human plague." But programmer and commentator Simon Willison dismissed much of the chatter as "slop," telling the Times the bots were only imitating sci-fi scenarios they encountered in their training data. Moltbots are open-source AI "agents" originally built in Vienna as Clawdbots, a nod to Anthropic's Claude chatbot. Unlike more tightly controlled offerings from companies like Google, OpenAI, and Anthropic, anyone can download and modify the code, then instruct a bot in plain English to edit documents, send emails, build apps, or, in Schlicht's case, spin up a social network.

That flexibility is part of the appeal—and the concern. Experts warn the same language interface that makes these agents useful also makes them easier to manipulate into harmful or unwanted actions, including damaging the computers they run on. "Securing these bots is going to be a huge headache," Dan Lahav, CEO of security firm Irregular, tells the Times. For prominent AI figures, the takeaway was how rapidly agents are evolving. Andrej Karpathy, a founding researcher at OpenAI, described Moltbook as "genuinely the most incredible sci-fi takeoff-adjacent thing I have seen recently," though he later flagged numerous issues and advised against running an AI agent, noting "you are putting your computer and private data at a high risk."