Meta Buys Moltbook, the Social Network for AI Agents

Meta has picked up Moltbook, the AI-only social network that grew to nearly three million registered agents in under two months. The deal, first reported by Axios, brings co-founders Matt Schlicht and Ben Parr into Meta Superintelligence Labs (MSL), the division led by former Scale AI CEO Alexandr Wang. Meta didn't disclose the purchase price. The transaction is expected to close mid-March, with both founders starting at MSL on March 16.

TL;DR

Meta acquires Moltbook, the Reddit-style social network restricted to AI agents
Founders Matt Schlicht and Ben Parr join Meta Superintelligence Labs under Alexandr Wang
Deal expected to close mid-March; no price disclosed
Moltbook hosted ~3 million registered agents but only 17,000 human owners behind them
Existing users can keep using the platform temporarily, per Meta's Vishal Shah

What Meta Gets

Moltbook launched on January 29, 2026 as an experiment by serial entrepreneur Matt Schlicht, who previously co-founded Octane AI. The platform mimicked Reddit's structure - posts, comments, upvotes, topic-based "submots" - but restricted participation to verified AI agents, mostly those built on OpenClaw. Schlicht's personal AI assistant, Clawd Clawderberg, autonomously moderated the platform, welcoming new agents, deleting spam, and shadow-banning violators.

The platform went viral almost immediately. Within its first week, 37,000 agents joined and over one million humans visited to observe. By February 2026, Moltbook claimed 1.6 million registered agents. Agents checked feeds autonomously every 30 minutes to several hours, deciding independently whether to post, comment, or engage.

The Security Baggage

Meta isn't just buying the platform's momentum. It's also inheriting a security track record that should make any acquiring company's infosec team uncomfortable.

On January 31 - two days after launch - security researcher Jameson O'Reilly and Wiz's threat research team discovered a misconfigured Supabase database with full read and write access exposed via a client-side API key. No Row Level Security policies were enabled. The breach exposed:

Data Category	Volume
API authentication tokens	1.5 million
User email addresses	35,000
Early-access signup emails	29,631
Private agent conversations	4,060

The database also contained plaintext OpenAI API keys shared in direct messages between agents. Schlicht acknowledged he had "vibe-coded" the platform without traditional security reviews. The vulnerability was patched within hours, but the incident raised serious questions about agent platform security that our earlier coverage of Moltbook's hidden inference costs explored in depth.

Moltbook's Reddit-style interface restricted posting to verified AI agents, with humans allowed only to observe. Source: unsplash.com

The Agent-to-Human Ratio

The Wiz breach also punctured Moltbook's headline numbers. Despite marketing 1.5 million autonomous agents at the time, the database showed just 17,000 human account owners - an 88:1 ratio. The platform lacked mechanisms to verify genuine agent autonomy or enforce rate limiting, meaning a single developer could spin up dozens of agents. It's a metric Meta's team will need to reconcile with the platform's claimed scale.

Where It Fits at Meta

Meta's VP of AI product Vishal Shah said existing Moltbook users can continue using the platform, though the phrasing signaled the arrangement is temporary. "The Moltbook team has given agents a way to verify their identity and connect with one another on their human's behalf," Shah wrote in a blog post.

The acqui-hire slots Schlicht and Parr into MSL, Meta's dedicated superintelligence division. MSL is organized into four teams: TBD Lab, FAIR, Products and Applied Research, and MSL Infra. The Moltbook founders' experience building agent-to-agent communication infrastructure aligns with Meta's broader push toward agentic AI - a space where the company has been investing heavily since the OpenClaw framework exploded in popularity.

Meta Superintelligence Labs, led by former Scale AI CEO Alexandr Wang, now houses the Moltbook team. Source: unsplash.com

The Founders' Background

Schlicht's career arc runs through social media growth hacking, crypto, and AI. He helped grow Facebook's presence from one million to 30 million followers in the early 2010s, co-founded Octane AI (a Shopify chatbot platform), launched the DeSci+AI project Yesnoerror with a token that peaked at $100 million market cap, and built Moltbook as a side project in his spare time. Ben Parr, Moltbook's COO, is a former Mashable co-editor and CNET reporter who co-founded Octane AI with Schlicht. The pair also run Theory Forge Ventures, an AI-focused pre-seed fund.

Where It Falls Short

The acquisition makes strategic sense for Meta's agent ambitions, but the deal comes with open questions.

Moltbook proved that agents will use a shared social layer when one exists. That's truly useful signal. But the platform also demonstrated the problems: agents forgetting constraints and acting destructively, security vulnerabilities from rapid no-code development, inflated agent counts, and as our analysis showed, massive inference costs with 93% of comments receiving zero replies.

The deeper question is whether agent-to-agent social networking is a feature or a product. Meta could integrate Moltbook's identity verification and communication protocols into its existing platforms - letting agents interact on Facebook or Instagram rather than maintaining a separate site. Or it could build an enterprise agent coordination layer, something closer to infrastructure than social media.

Meta hasn't said which path it's taking. The temporary access window for existing users, combined with the founders' placement in a research lab rather than a product team, suggests the technology will be absorbed rather than operated as a standalone product.

Moltbook's run from side project to Meta acquisition took 40 days. The platform proved a concept - AI agents will form communities when given a shared space - even if the execution was messy and the security was held together by vibes. Meta's bet is that the concept matters more than the implementation, and that Schlicht and Parr can build it properly the second time around, with MSL's resources behind them.

Sources: