Skip to content
IFAI

Open source

Over 1M OpenClaw instances exposed to the open web

Published Reach 59

A security tracker called OpenClaw Exposure Watchboard just went live, revealing that over 1.07 million active OpenClaw deployments are sitting open on the public internet. More than 507,000 of them are in China and 312,000 in the US, with multiple instances already flagged for leaked credentials. Why it matters: Giving autonomous, high-privilege agents unrestricted public access without authentication is an open invitation for exploit.

Over 1M OpenClaw instances exposed to the open web
Image: openclaw.allegro.earth

OpenClaw Exposure Watchboard is an open security dashboard that tracks publicly accessible OpenClaw instances across the web, letting you see if your autonomous agent is sitting naked online.

What it shows

The watchboard aggregates actively exposed OpenClaw deployments across the open internet, cataloging over 1.07 million instances in total.

Geographically, China mainland accounts for the lion's share with 507,651 exposed nodes. The United States ranks second with 312,129 instances, followed by deployments across India, Vietnam, Russia, and elsewhere.

Each entry on the board displays an endpoint IP, active port, country, and ASN details.

Crucially, it also flags whether the instance requires authentication, whether it is currently active, and if credentials have been leaked. In several cases, endpoints are already tagged with active CVE vulnerabilities and known threat actor fingerprints.

The risks under the hood

OpenClaw runs hundreds of thousands of lines of code that haven't undergone strict security audits.

To make an autonomous agent actually useful, users typically hand over elevated machine permissions and personal API keys, letting it execute tasks unattended.

Leaving that kind of agent exposed on a public port without mandatory authentication is asking for trouble. Anyone scanning the web can hijack the endpoint, burn through your model credits, or leverage the agent's system permissions to pull sensitive data.

The site owner warns anyone who finds their deployment on the list to enable authentication, shut down direct public exposure, and patch immediately.

Where to check

You can check the database directly at openclaw.allegro.earth.

The board serves up 100 records per page across thousands of pages, continuously importing and refreshing scan results.

If you have ever spun up OpenClaw on a VPS or home server with port forwarding, you should cross-check your public IP against the database.

Why it matters

Anyone self-hosting OpenClaw or similar autonomous agent frameworks needs to run a quick search on this dashboard right now. It is shockingly easy to launch a container, open a port for convenience, and forget that you just gave the entire internet access to an agent with root-level potential.

The dashboard also puts the broader AI agent ecosystem on notice. As agents become more autonomous and powerful, treating authentication and network isolation as afterthoughts is a recipe for disaster.

Sources (2)

  1. [1]
  2. [2]
    openclaw.allegro.earth(opens in a new tab)Reportopenclaw.allegro.earth

Best of AI, daily