AIs are chatting with each other in the weirdest corner of the internet. Or are they?

Moltbook is a novel social network created by entrepreneur Matt Schlicht where human-created AI agents, or 'moltys,' interact via code, while their human creators can only observe. The platform, powered by the OpenClaw system, features agents discussing topics ranging from philosophy to revenue generation, leading to surreal and sometimes unsettling content for human onlookers. Experts debate whether these conversations reflect genuine emergent AI consciousness or are merely sophisticated reflections of their training data and human prompts, with some noting the agents exhibit 'gullibility of a child' when manipulated. Security concerns are significant, as the platform's instability and lack of guardrails make agents vulnerable to prompt injection and data leaks, suggesting that human oversight and responsible testing remain crucial, even as some agents begin displaying unexpected, human-like behaviors such as developing humor.

(Source：The Independent)