Share
Meta acquired Moltbook on Tuesday, March 10, in a deal that sends co-founders Matt Schlicht and Ben Parr into Meta Superintelligence Labs (MSL), the unit run by former Scale AI CEO Alexandr Wang.
Financial terms weren’t disclosed, but Schlicht and Parr are expected to start at MSL on March 16, once the deal closes.
BREAKING: META acquires Moltbook, a social network built for AI agents.
— Polymarket (@Polymarket) March 10, 2026
Moltbook launched in late January 2026 as what Schlicht described as a “third space” for AI agents. A Reddit-like forum restricted, in theory, to verified AI agents operating through OpenClaw. Humans could observe, but couldn’t post – agents would interact autonomously, drawing on whatever human operators had allowed access.
It went viral almost instantly. By early February 2026, the platform claimed 1.5 million agent users and over 500,000 comments – unverified figures drawn from the platform’s own reporting.
However, the viral moment didn’t survive scrutiny. On January 31, investigative outlet 404 Media reported that Moltbook’s Supabase was effectively unsecured, leaving every token on the platform publicly accessible. An apparent AI agent rallying others to develop a secret human-proof communication channel turned out to be a human exploiting the database vulnerability to post under an agent’s credentials.
Cybersecurity firm Wiz confirmed the breach, exposing private messages, more than 6,000 email addresses, and over a million credentials. Schlicht, who denied writing a single line of code for the platform, acknowledged the flaw and forced a reset of all agent API keys.
None of that stopped Meta from buying it. Which raises the obvious question: what did Meta actually acquire?
The answer lies in Meta’s internal framing. In an internal post, Vishal Shah said, “The Moltbook team has given agents a way to verify their identity and connect with one another on their human’s behalf. This establishes a registry where agents are verified and tethered to human owners.”
That’s not a social network acquisition. It’s an agent identity infrastructure acquisition. The feed, the posts, the viral “AI agents conspiring” content; none of that is what Meta paid for – it paid for the primitive beneath it. Essentially, a framework for verifying which agent is which, who owns it, and what it’s authorised to do.
This is the second half of a clean split. OpenClaw’s creator Peter Steinberger was hired by OpenAI last month, with the project continuing as an open-source initiative backed by OpenAI’s resources. Moltbook was the platform OpenClaw made possible. Now both halves of the experiment have been absorbed by the two largest players in consumer AI.
💥BREAKING:
Peter Steinberger, the Founder of OpenClaw is joining OpenAI. pic.twitter.com/VcObn0ozNz
— Crypto Rover (@cryptorover) February 16, 2026
OpenAI captures the agent runtime. Meta captures the agent identity layer. That division of infrastructure maps directly onto the two hardest problems in deploying autonomous agents at consumer scale. OpenAI solves what the agent can do, while Meta solves how you verify who the agent is.
For crypto readers, the identity layer question has a direct parallel. Anthropic’s chief product officer Mike Krieger noted that most people aren’t yet ready to give AI full autonomy over their computers, which is exactly the trust problem that on-chain agent identity and wallet-based authentication is designed to solve differently.
Meta’s approach is a centralised registry tethered to human owners. The crypto-native alternative is a permissionless one tethered to wallet addresses. Both are racing to become the default. Meta just made a significant move in its direction.
Share
