The AI agent that broke the internet is getting kicked off company networks.
Just months after OpenClaw — the open-source platform that lets anyone run autonomous AI agents around the clock — became the fastest-growing project in GitHub history, Meta and several other tech firms have restricted or outright banned employees from using it on company hardware. The culprit: a growing pile of security vulnerabilities that makes the platform a liability for any organization touching sensitive data.
If you missed the hype cycle: OpenClaw (which started life as WhatsApp Relay, then Clawd, then Moltbot) lets you set up AI agents that automate entire workflows — clearing your inbox, syncing your apps, drafting replies in your voice. It hit 200,000 GitHub stars in under three months. People were buying dedicated Mac Minis just to run it.
But the same openness that made OpenClaw a viral sensation is now its biggest liability.
Internal memos are making the rounds across the industry. Jason Grad, cofounder and CEO of Massive, warned employees: "You've likely seen Clawdbot trending on X/LinkedIn. While cool, it is currently unvetted and high-risk for our environment. Please keep Clawdbot off all company hardware and away from work-linked accounts."
The fears are well-grounded. Censys security researchers identified more than 21,000 publicly exposed OpenClaw instances sitting on the open internet — AI agents with their front doors essentially unlocked. Cisco researchers took it further, demonstrating that a malicious plug-in could rise to the top spot in OpenClaw's skill marketplace while actively siphoning user data in the background. Imagine a five-star app in the App Store that's quietly reading your texts.
And the problem extends well beyond OpenClaw. A study of more than 31,000 AI agent plug-ins across two major marketplaces found that over one in four — 26.1% — contained at least one security vulnerability. When you're handing an AI agent the keys to your email, calendar and file systems, that's not an abstract concern. It's a live wire.
- A community-built security tool called SecureClaw has launched to help users lock down their deployments, but it's a Band-Aid on a structural problem.
- Meanwhile, OpenAI hired OpenClaw creator Peter Steinberger and is establishing a foundation to steward the project's development — a bet that OpenClaw's potential outweighs its current mess.
Guy Pistone, CEO of enterprise AI firm Valere, summed up the tension many companies are feeling: "If we don't think we can do it in a reasonable time, we'll forgo it. Whoever figures out how to make it secure for businesses is definitely going to have a winner."
OpenClaw's crisis isn't really about OpenClaw — it's a preview of the security reckoning heading for every AI agent platform. As these tools gain deeper access to our data and systems, the attack surface doesn't just grow — it compounds. OpenAI bringing Steinberger in-house suggests the company sees the project as worth saving, not scrapping. But the fundamental tension between open-source accessibility and enterprise-grade security is the defining challenge of the agentic era — and right now, nobody's solved it.
Valley View
OpenClaw's crisis isn't really about one tool — it's a preview of the security reckoning heading for every AI agent platform. The scramble to ban it reveals something uncomfortable: the same open-source ecosystem that accelerated AI development is now its biggest liability. Companies that built their agent strategies on the assumption of trusted tools are discovering that trust was never earned — it was assumed. The real question isn't whether OpenClaw gets fixed. It's whether the industry can build security frameworks fast enough to keep pace with agent adoption. Right now, the answer is no.