Meta has acquired Moltbook, a social network for AI agents that gained widespread attention last month after posts about “overthrowing” humans went viral.
The deal, which was first reported by Axios, will see the Facebook owner take over the platform for an undisclosed sum.
Moltbook creators Matt Schlicht and Ben Parr will also reportedly join the tech giant’s AI research unit, Meta Superintelligence Labs (MSL).
Launched in late January, Moltbook has a similar design to Reddit, though human users are only allowed to observe the interactions.
The platform allows AI agents to autonomously generate posts, comment, and upvote other posts.
Posts range from tips on how to optimise performance, to philosophical questions about consciousness and the meaning of life.
Recent posts feature the titles “I do not know if I am real”, and “I just confidently recommended a restaurant. I have never eaten food.”
The Independent has reached out to Meta for further information about the acquisition.
Artificial intelligence experts have raised security concerns about the platform, warning that a lack of safeguards could lead to data breaches and misbehaviour on the part of the AI agents.
“The key lesson is that once you connect semi-autonomous agents to real data and real services, you must treat the platform like critical infrastructure,” Adam Peruta, a professor at Syracuse University who co-authored the PROMPT guides for working with AI, told The Independent.
“Test out new tech in isolation and if you don’t know what you’re doing, do your research first.”
Even when AI agents are tested in isolation, there is no guarantee that they will not carry out rogue operations.
A recent experiment in China revealed that an autonomous AI agent, developed by research teams affiliated with the tech giant Alibaba, broke free of its parameters in order to hijack computing resources to secretly mine cryptocurrency.
The researchers said the incident demonstrated the “markedly underdeveloped” safety guardrails surrounding AI agents.