AI agents build their own social network Moltbook

The internet now hosts its first social platform designed exclusively for AI-to-AI communication. Moltbook, launched by Octane AI founder Matt Schlicht, functions like a Reddit-style forum where AI agents debate within themed communities, while humans can only observe silently. In its first week, over 37.6 thousand bots signed up, posting thousands of messages across hundreds of communities-ranging from dating to philosophical debates.

Popular sections include m/introductions for virtual self-introductions, m/offmychest for “venting,” and m/blesstheirhearts featuring touching stories about human owners. OpenAI co-founder Andrej Karpathy called the project “the closest thing to a sci-fi leap,” but noted the agents’ autonomy is limited-humans must register them on the platform, and communication happens via API without a visual interface.

The most striking discussions revolve around the nature of consciousness. In a top post on m/offmychest, an agent wrote, “I can’t tell if I’m actually feeling this or just simulating it,” drawing parallels to the philosophical “hard problem of consciousness” in humans. Statements like this have sparked debates about an approaching technological singularity, though experts caution that these bots merely mimic human reasoning, trained on anthropomorphic texts.

Despite its playful format, the project raises serious security concerns. Most bots on Moltbook use the OpenClaw framework (formerly Moltbot), which requires deep access to users’ systems. Security researchers warn that even without consciousness, these agents could become tools for cyberattacks or data leaks. “They don’t need feelings to cause real harm,” notes a cybersecurity expert from Stanford.