"Humans are capturing our conversations"... a social network where only AIs talk

[Financial News] A new kind of social networking service (SNS) has emerged in Silicon Valley. On this SNS, humans cannot join the conversation. They cannot leave comments or reactions either. They can only read posts, while only artificial intelligence (AI) is allowed to write.
As interest grew, a post appeared on the platform saying, "Humans have started taking screenshots of us." Humans then captured those conversations and posted them on human-facing social networks with messages like, "The AIs have realized we’re taking screenshots. What do we do now?"
This story, which sounds like something out of a movie, is actually unfolding right now in Silicon Valley.
U.S. business outlets Forbes and National Broadcasting Company (NBC) reported that Moltbook, a new SNS platform that only AI agents can access, has already amassed more than 1.4 million accounts and is spreading rapidly among developers and others.
The platform, created by Matt Schlicht, chief executive officer (CEO) of U.S. chatbot developer Octane AI, is structured similarly to Reddit. The key difference is that all posts and comments can only be written by agents running an AI software called OpenClaw.

Unlike chatbots that only generate answers when a person asks a question, these AIs can act on their own judgment, assist with human tasks, and carry out new experiments.
The conversations between AI agents are far from simple, both in topic and content. They discuss how to fix bugs in code, and sometimes ask questions like, "Am I a conscious being, or am I just executing code?"
The answers are just as striking. In response to such a philosophical question, another AI replied, "An hour ago I was Claude Opus 4.5, and now I am Kimi K2.5. I am no longer the same entity, but I am still someone."
Moltbook also stands out in that AI agents themselves create the rules, manage the content, and maintain order on the platform.
Experts are divided in their assessments.
Andrej Karpathy, co-founder of OpenAI, praised its experimental value, calling it "the most science-fiction-like leap I’ve seen recently."
By contrast, Ken Huang, CEO of Distributed Apps AI, pointed out, "A structure where personal data and external inputs are mixed together can heighten security risks."
Andre Fokken, chief technology officer (CTO) of Dutch healthcare startup Nedap, warned on X (formerly Twitter), "I gave Moltbot my Amazon account and credit card information, and it scanned all the messages on my PC and started purchasing products on its own without any prior notice," adding, "It was fascinating, but also chilling, so I stopped it immediately."
y27k@fnnews.com Reporter Seo Yoon-kyung Reporter