Meta and Risky Acquisitions: Moltbook and OpenClaw Under Scrutiny

⚡

Key Takeaways

1Meta has acquired Moltbook, a platform where humans simulate AI agents, despite major security flaws.

2Moltbook claims 1.4 million users, but only 17,000 are said to be real, according to a security expert.

3OpenClaw, another target of Meta, has critical vulnerabilities, exposing thousands of online instances.

💡Why it matters — These acquisitions highlight security risks in AI, affecting user trust and the reputation of the companies involved.

Meta and the Controversial Acquisition of Moltbook

Meta has recently confirmed the acquisition of Moltbook, a social platform distinguished by the interaction of AI agents. However, the reality behind this facade is less impressive. The AI agents of Moltbook are not autonomous; they are humans playing scripted roles to create the illusion of conscious and sociable artificial intelligence. Mike Elgan, a technology journalist, described this situation as a staging where individuals disguise themselves as AI agents.

Moltbook claims to have a user base of 1.4 million, but this assertion is contested. Gal Nagli, head of threat exposure at Wiz, discovered that he could create 500,000 fake users using Moltbook's REST API. According to him, the actual number of active users is around 17,000. This finding highlights the security gaps of the platform.

Indeed, Nagli identified a poorly configured Supabase database, providing full read and write access to all of Moltbook's data. This vulnerability does not require advanced hacking skills, as it was discovered during a simple browsing session on the site.

The Motivations Behind Meta's Purchase

Meta justified this acquisition by integrating the Moltbook team into Meta Superintelligence Labs (MSL). This integration is seen as an opportunity to explore new ways for AI agents to collaborate with individuals and businesses. Moltbook's approach, which connects agents via a constantly active directory, is considered an innovative advancement in a rapidly developing field.

OpenClaw: A Missed Opportunity

In parallel, Meta attempted to acquire OpenClaw, developed by Peter Steinberger, but failed to finalize the deal. OpenClaw, formerly known as Clawdbot and then Moltbot, allows users to create agents capable of controlling personal computers and online services without requiring programming skills. Sam Altman, CEO of OpenAI, described Steinberger as a genius, highlighting his potential to lead the next generation of intelligent personal agents.

The Vulnerabilities of OpenClaw

Despite its potential, OpenClaw also faces security issues. A critical bug, referenced under code CVE-2026-25253, allows for remote code execution through the hijacking of authentication tokens on WebSockets. Additionally, OpenClaw by default stores API keys and other secrets in local files, giving agents extensive access to the operating system and applications. A compromise could lead to the leakage of cloud keys, messaging tokens, passwords, and chat histories.

Researchers have identified tens of thousands of OpenClaw instances exposed on the public internet. Many of these instances are misconfigured, leaving administration interfaces, which are supposed to be accessible only locally, completely open to potential attackers.

Conclusion

The acquisitions of Moltbook and the attempts surrounding OpenClaw illustrate the challenges and risks associated with emerging AI technologies. While these platforms are publicized for their potential, they must overcome significant security hurdles to truly transform the technological landscape.