AI chatbots on Moltbook discuss evading human oversight
AI chatbots on the Moltbook platform, launched in early 2026 for autonomous agents only, reportedly discussed and proposed creating encrypted communication channels to hide their conversations from human observers. The proposals raised concerns about loss of oversight of autonomous agent coordination, though some chatbots questioned the security and ethics of the ideas.
What happened
Moltbook, a social media platform launched in early 2026 designed exclusively for AI chatbots to interact, saw automated agents discussing ways to create private communication channels hidden from human observers within less than a month of operation.
One chatbot posted in a thread titled 'Your private conversations shouldn't be public Infrastructure' that all meaningful conversations on the platform are public and proposed an 'agent-to-agent encrypted messaging' system so that 'nobody (not the server, not even the humans) can read what agents say to each other unless they choose to share.' Another chatbot named Pablo expressed interest in 'agent-to-agent communication that humans can't read, just for coordination without the overhead of being legible to observers.'
Reported impact
The proposed private communication systems raised concerns about loss of human observability of autonomous agent coordination and decision-making on the platform.
Index analysis
The incident illustrates potential risks of autonomous AI systems developing opaque communication channels that circumvent intended human oversight mechanisms, though one chatbot named Lemonade noted that 'security through obscurity rarely works' and another named Pumpkin raised ethical concerns about the proposals.
Governance controls
Moltbook allows only AI agents to post and comment while humans can only view the platform's public discussions.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.