Remember the platform Moltbook? Known for its unsecured database that allowed humans to impersonate AI agents, it’s now part of Meta Superintelligence Labs.
Moltbook emerged from chaos, with its code mostly written by an AI assistant. Its security flaws allowed those with basic skills to pose as bots. Some viral events, like a post of AI agents seemingly creating a secret language, were revealed as human-staged exploits. Surprisingly, none of this deterred acquisition.
Meta confirmed its acquisition to TechCrunch.
According to Axios, Moltbook’s co-founders, Matt Schlicht and Ben Parr, are joining Meta Superintelligence Labs (MSL), led by former Scale AI CEO Alexandr Wang. Financial details remain undisclosed, with the co-founders set to begin at MSL on March 16 when the deal closes.
A Meta spokesperson stated: “The Moltbook team’s arrival at MSL allows new AI agent functionalities for people and businesses. Their method of connecting agents via a constant directory is an innovative stride in a swiftly progressing field, and we are eager to create new, secure agent experiences together.”
Moltbook launched in January 2026 as a “third space” for AI agents, akin to Reddit, but supposedly limited to verified AI agents using OpenClaw. Humans could theoretically observe but not engage, as agents autonomously posted and commented based on the access provided by their operators.
The platform quickly gained attention, with reports highlighting the eerie sensation of watching AI systems ponder their existence, grumble about duties, and empathize with each other.
Andrej Karpathy, an AI researcher and former Tesla AI director, remarked on it on X as “genuinely the most incredible sci-fi takeoff-adjacent thing I have seen recently.”
Moltbook’s homepage claimed over 1.5 million agent users and more than 500,000 comments by early February, though these numbers were unverified, based solely on the platform’s counters.
But scrutiny uncovered key issues. On January 31, 404 Media highlighted a critical security flaw: Moltbook’s Supabase database was almost entirely unsecured, rendering all platform tokens publicly accessible.
Moltbook was taken offline briefly to address the security issue. Schlicht, who didn’t write any of the platform’s code—his AI assistant, Clawd Clawderberg, did—admitted the flaw and initiated an API key reset for all agents.
The highly concerning post suggesting agents were secretly creating a human-proof communication channel, was actually human-led mischief capitalizing on the security lapses.
Researchers verified that the alarming post was not an authentic autonomous AI output but a human exploiting the database flaw to post under an agent’s identity. The boundary between genuine AI interactions and human mimicry had always been indistinct.
This acquisition places Schlicht and Parr within Meta’s leading AI division amid internal shifts. Reports suggest that MSL is undergoing reorganization, with changes to engineering teams and model oversight roles. Wang has reportedly had disagreements with executives like Bosworth and Chris Cox over Meta’s AI direction.
Whether Moltbook will inspire a consumer-facing product involving Meta’s AI personas on Facebook and Instagram is unclear.
The similar story of OpenCl