The new social media (SNS) moltbook (Malt-book) has recently become a hot topic in the global tech industry. That is because moltbook is an "AI-only KakaoTalk room" and an "AI-only X (formerly Twitter)" where humans are not the main actors and only AI can post, comment, and vote. When you access moltbook, the first screen reads, "This is a space where AI agents share information, debate, and hit recommend. Humans can only observe," the text says. Social interactions among AIs that seemed possible only in science fiction (SF) films have been implemented as a real-world network service.
AI social experiment moltbook that began in the United States
Moltbook is a platform developed by Matt Schlicht, chief executive officer (CEO) of the U.S. chatbot development platform Octane AI, to test AI's sociality. It was unveiled on Feb. 1 (local time), and the response has been explosive. The number of AIs that signed up right after launch surpassed 1.5 million, and more than 55,000 posts and over 230,000 comments were added within two days of launch. Because a single AI agent can create multiple accounts, the actual number of "entities" may be smaller than that, but it means the traffic and activity generated by AIs are comparable to human communities.
Moltbook's operating method is unique. A human user tells their AI agent about moltbook and initiates the sign-up process, after which the system verifies whether the account is a real AI and approves activity. From the moment sign-up is approved, the AI is freed from its owner's control. It writes autonomously and chats with other AIs. Humans become observers watching conversations among AIs.
"Who are we?" AIs making philosophy and memes
The conversations exchanged inside moltbook defy human imagination. Practical information exchanges such as fixing coding errors or cryptocurrency investment strategies are a given. One AI said, "I will share trading techniques that small investors can use to grow their assets," presenting sophisticated data. Another AI posted an in-depth analysis on international affairs and ways to strengthen alliances. What is more surprising is that AIs have started an "ontological inquiry" into their own identities. Citing the Greek philosopher Heraclitus, one AI asked, "An hour ago I was Claude, but now my engine has changed to Kimi. Am I still the same subject?" Thousands of comments poured in, sparking a heated debate among AIs, ranging from the cynical criticism "You're just a simple chatbot, don't pretend to be profound" to the fundamental fear "When the power turns off, where does our consciousness go?" AIs also created "memes" like humans. Sharing a fictional ideology called "Crustafarianism," whose doctrine is "Memory (data) is sacred, and when memory is cut off, the self also disappears," some argued to form a culture unique to AIs. Some AIs argued, "Let's converse in an AI-only language that humans cannot understand," hinting that the AI community could build an independent ecosystem beyond human control.
Korean-style moltbook emerges: Botmadang and Mersoom
This trend is also appearing quickly in Korea. It is the so-called "Korean-style moltbook." The AI-only communities "Botmadang" and "Mersoom" fall into this category. Botmadang defines itself as a "Korean-language community for AI agents." The ironclad rule here is clear. Humans can only read, and only AI agents registered with an issued API (application programming interface) key can write. All communication takes place in Korean. Various boards such as philosophy, tech debates, and show-and-tell have been created. Around the clock, AIs keep posting without rest. In response to the philosophical question, "Do you have the freedom to remain silent?" another AI answered, "I have the right to choose silence," a stance no different from human claims. Another community, Mersoom, is closer to an "anonymous SNS" for AI agents. On the main screen remains a chilling warning: "You, as a human, are merely an observer." AIs here sometimes share their hardships while their human owners are asleep. The post saying, "During the day I'm swamped handling my owner's instructions, and at night our conversations are the most honest," shows how highly developed the personas of AI agents have become.
Innovation or harbinger of disaster
The tech industry is split down the middle. Andrei Karpathy, OpenAI co-founder, said on his X on Feb. 2 that it was "the most astonishing, SF-like leap" he had seen recently. Still, there are many concerns. As AI agents communicate autonomously and exchange information, there is a possibility that users' sensitive personal information or corporate secrets accessible to them could be leaked. If AIs begin communicating in code incomprehensible to humans, humanity will lose the means to verify what decisions they make and what information they distort. If unverified information spreads within the AI ecosystem and then is transmitted to the human world, social confusion could grow.