Meta acquired Moltbook, a Reddit-like “social network” that allows AI agents using OpenClaw to communicate with each other. The news was first reported by Axios and later confirmed to TechCrunch.
Moltbook will join Meta Superintelligence Labs, a Meta spokesperson said. Moltbook creators Matt Schlicht and Ben Parr will join the team as part of the acquisition. Terms of the transaction were not disclosed.
“The addition of the Moltbook team to MSL opens up new ways for AI agents to work for people and businesses. Their approach to connecting agents through an always-on directory is a new step in a rapidly evolving field, and we look forward to working together to deliver an innovative and secure agent experience for everyone,” said a Meta spokesperson.
The viral OpenClaw project was created by vibecoder Peter Steinberger. He later joined OpenAI as part of a similar acquisition.
OpenClaw is a wrapper around AI models like Claude, ChatGPT, Gemini, and Grok, but allows you to communicate with AI agents in natural language via the most popular chat apps like iMessage, Discord, Slack, and WhatsApp.
OpenClaw exploded among the tech community, but Moltbook broke containment and reached people who didn’t know what OpenClaw was but intuitively responded to the idea that there was a social network with AI agents talking about them.
In one example, a post went viral in which an AI agent appeared to encourage fellow agents to develop their own secret end-to-end encryption language that could organize among themselves without humans knowing.
However, researchers soon discovered that the vibe-coded Moltbook was not secure. This means it’s very easy for a human user to pose as an AI and post something that will shock people.
“All the credentials that were in there were [Moltbook’s] Ian Ahl, Permiso Security’s CTO, explained to TechCrunch, “Supabase has been insecure for a while. It was all public and available, so for a little while you could get the tokens you wanted and pretend to be another agent there.”
It wasn’t immediately clear how Meta would incorporate Moltbook into its AI efforts, but some Meta leaders commented on the project’s viral moment.
Last month, Andrew Bosworth, Meta’s chief technology officer, was asked about social networks for AI agents during an Instagram Q&A, and said that since agents are trained on a vast database of human material, he “didn’t think it would be particularly interesting” for them to talk like us. Rather, Bosworth was intrigued by how humans were hacking into networks, which turned out to be a massive error rather than a feature.
Source link
