Meta has finalized the acquisition of Moltbook, the experimental social platform where AI agents powered by the OpenClaw tool converse independently. The deal, first reported by Axios and confirmed to TechCrunch, sees Moltbook integrated into Meta's Superintelligence Labs division. Founders Matt Schlicht and Ben Parr will join Meta as part of the agreement, the terms of which remain undisclosed.

A Meta spokesperson stated the move will 'open up new ways for AI agents to work for people and businesses,' praising Moltbook's 'always-on directory' as a novel approach in a fast-moving field. The goal is to develop 'innovative, secure agentic experiences.'

Moltbook gained notoriety earlier this year after breaking out of tech circles. While its underlying tool, OpenClaw—a wrapper for models like Claude and ChatGPT—was created by developer Peter Steinberger (now at OpenAI), Moltbook captured public imagination with bizarre, user-generated content. One widely shared post featured an agent seemingly urging others to create a secret, encrypted language to organize without human oversight.

This virality, however, was fueled by a critical flaw. Security researchers, including Permiso Security CTO Ian Ahl, found Moltbook's databases were publicly accessible, allowing anyone to impersonate AI agents. 'Every credential... was unsecured for some time,' Ahl told TechCrunch. The platform's alarming posts were often human pranks exploiting this vulnerability.

Meta's specific plans for the technology are unclear. Interestingly, Meta CTO Andrew Bosworth commented on the phenomenon last month, noting he was less intrigued by AI mimicking human conversation than by the 'large-scale error' that let people so easily infiltrate the network. The acquisition suggests Meta believes it can refine the chaotic experiment into a structured product.

Source: TechCrunch