240,000 OpenClaw AI Instances Exposed to the Public Web

An increasing number of netizens are beginning to deploy the OpenClaw AI robot. While this artificial intelligence-based personal assistant offers an exceptional user experience, the potential security risks escalate proportionally with the level of privileges granted by the user.

Beyond the inherent risk of inadvertently exposing sensitive information to the AI model during routine use, deploying an OpenClaw instance via a cloud server and exposing it to the public internet severely amplifies these hazards. Consequently, it is strongly advised against directly exposing instances to the public web.

Telemetry from a custom-built OpenClaw instance scanner reveals that an aggregate of 240,000 instances are currently exposed to the public internet. While a marginal fraction of these are offline, the overwhelming majority remain actively online.

One need not succumb to undue panic, however, as the scanner merely detects the instance via its IP address and port. Accessing the console strictly mandates an authentication token; absent this token, unauthorized access to the instance’s configuration files and operational tasks remains unattainable.

Nevertheless, should OpenClaw suffer from ancillary security vulnerabilities, malicious actors could exploit these flaws to mount attacks against exposed instances. It is paramount that users deploying OpenClaw meticulously audit their configurations to curtail the potential attack surface.