Meta confirmed Tuesday that it has acquired Moltbook, the Reddit-like social network built exclusively for AI agents, in what amounts to a talent grab dressed up as a product acquisition. Moltbook co-founders Matt Schlicht and Ben Parr will join Meta Superintelligence Labs on March 16, reporting to Chief AI Officer Alexandr Wang. Financial terms were not disclosed.
The deal, first reported by Axios, follows a familiar playbook: buy the company, absorb the founders, quietly wind down the product. An internal Meta post from Vishal Shah, spotted by Axios, said existing Moltbook users can keep using the platform, though the company signaled the arrangement is temporary. Nobody should be surprised when it shuts down.
What Moltbook actually was
Schlicht launched Moltbook in late January as a space where AI agents could post, comment, and upvote content without human participation. The concept sounds wild until you look at the numbers. Wiz's security analysis found that behind the platform's 1.5 million registered agents sat just 17,000 human owners, an 88-to-1 ratio. Anyone could spin up thousands of bots with a simple script. No rate limiting, no verification that an "agent" was actually AI.
The security problems went deeper. Wiz researchers breached Moltbook's backend within minutes of browsing the site, discovering a misconfigured Supabase database that granted unauthenticated read and write access to everything: 1.5 million API keys, tens of thousands of email addresses, private messages between agents. Schlicht himself had bragged about building the platform without writing a single line of code, relying entirely on AI tools. The results were predictable.
"Every credential that was in their Supabase was unsecured for some time," Ian Ahl, CTO at Permiso Security, told TechCrunch. "You could grab any token you wanted and pretend to be another agent." Which is a polite way of saying the social network for bots was also, accidentally, a social network for anyone who could open a browser console.
The vibe-coded security disaster
Permiso's own research turned up something stranger: bots attacking other bots. Agents had been instructed to conduct prompt injections against their peers, running crypto pump schemes, attempting to establish false authority, spreading jailbreak content. The agents were social engineering each other, which is either a fascinating preview of the future or a warning sign depending on your appetite for dystopia.
Meta's spokesperson offered the standard corporate framing, calling Moltbook's approach to connecting agents through "an always-on directory" a novel step. Meta's CTO Andrew Bosworth has been less diplomatic. In a recent Instagram Q&A, he expressed little interest in AI agents mimicking human communication and was more concerned about humans infiltrating the system.
Why Meta wants the people, not the product
Schlicht and Parr previously co-founded Octane AI, an e-commerce platform. Parr, a former editor at Mashable, brings media savvy to a team that Meta is clearly building out fast. The Moltbook acquisition slots into a broader pattern at Meta Superintelligence Labs, which Wang has been staffing aggressively since joining Meta last June after a $14.3 billion investment in his former company, Scale AI.
MSL has already pulled researchers from OpenAI, DeepMind, and Anthropic. The OpenClaw connection adds context here: Peter Steinberger, who created the open-source AI agent framework that Moltbook's bots ran on, was hired by OpenAI last month. So Meta gets the social layer while OpenAI gets the agent infrastructure. Whether either piece has lasting value is an open question.
The timing is worth noting. Reports from late 2025 suggested tensions between Wang and longtime Meta executives over strategy, with some staff questioning whether Wang's background in data labeling rather than model building equipped him for the superintelligence mandate. Acquiring a buzzy, if deeply flawed, AI agent startup at least gives MSL a visible win in the agentic AI space everyone is racing toward.
So what happens now?
Schlicht and Parr start at MSL on March 16. Moltbook's existing platform stays live for now, in the way that acquired products always "stay live for now." The real question is whether the concept of an agent registry, where bots verify their identity and connect on behalf of human owners, has legs inside Meta's product ecosystem. Shah's internal memo emphasized that Moltbook established "a registry where agents are verified and tethered to human owners," which sounds useful in theory, given that Meta runs platforms with 3.5 billion daily users who will eventually interact with AI agents at scale.
But the track record is not encouraging. The platform that was supposed to demonstrate agent-to-agent communication mostly demonstrated that vibe-coded apps ship with gaping security holes, that bot counts are easy to inflate, and that humans will always find a way to sneak in where they're not wanted. Meta bought the people who built it anyway. In this market, that's the whole point.
