You need to enable JavaScript to run this app.
Elastic Compute Service

Elastic Compute Service

Copy page
Download PDF
Managing application updates
Creating OpenClaw (Linux) task to deploy AI Assistant Security service
Copy page
Download PDF
Creating OpenClaw (Linux) task to deploy AI Assistant Security service
You can use the application update feature of Orchestration to batch enable the AI Assistant Security service for existing OpenClaw ECS instances, enhancing their security and controllability.​
You can also log in to an instance and enable the AI Assistant Security feature by executing relevant commands to obtain the corresponding security protection capabilities. Learn more.​
tip​
Currently, only the following regions are supported: China (Beijing), China (Shanghai), China (Guangzhou). Learn more.​
Background​
As a powerful AI agent, OpenClaw has the permission to directly operate devices, read local files, call system commands, access the network, and more. If OpenClaw is improperly configured or maliciously exploited, it can easily lead to security risks such as data leakage, system damage, and financial loss. To address such risks, Volcano Engine provides AI Assistant Security, which offers all-round security protection for OpenClaw and ensures its safe, stable, and reliable operation in your environment.​
AI Assistant Security acts as a dedicated security guardian for your AI assistant. Without affecting the normal operation of OpenClaw, it uses a built-in large model application firewall to monitor and defend against three core types of risks in real time:​
  • Prompt injection​
  • Attackers may hide malicious instructions in web pages, documents, or emails processed by OpenClaw, tricking it into performing unintended operations, such as stealing passwords or leaking data.​
  • PII leakage​
  • You may accidentally enter personal or corporate sensitive information (such as ID numbers, phone numbers, and API Keys) during AI conversations or document processing. This information may be sent to external large models, resulting in privacy breaches.​
  • Risky operations​
  • When interpreting ambiguous instructions, the AI agent may misinterpret instructions or be maliciously induced to call high-privilege tools and execute dangerous commands. For example, misinterpreting "clean up the desktop" as deleting critical system files.​
Prerequisites​
You have installed the cloud assistant Agent.​
Usage notes​
  • The security protection deployment does not affect the normal running of instances, requires no instance restart, and will not interrupt services.​
  • For an ECS instance with AI Assistant Security enabled, redeployment only performs an overwrite installation and updates the service to the latest version.​
Procedure ​
  1. Log in to the ECS console.​
  1. In the left-side navigation pane, click "Maintenance & Monitoring > Orchestration".​
  1. In the top menu bar, select the region and project of your business.​
  1. On the "CloudOps Orchestration" page, click the "Common O&M tasks" tab and select "App update".​
  1. Click the "Create task" button and configure the parameters described in the following table.​
  • ​
    Parameter​​
    Description​
    Application configuration​
    ​
    Application template​
    The OpenClaw application you deployed. The default name is "OpenClaw", which cannot be modified currently.​
    Task template​
    The preset standardized O&M process template. Select "Enabling AI Assistant Security", read and fully understand the "AI Assistant Security SDK Protocol", and then select "Agree AI Assistant Security SDK Protocol".​
    Select instances​
    ​
    Select instances​
    Select the instances for which you want to deploy security protection.​
    tip​
    By default, OpenClaw ECS instances are filtered by "Application template: OpenClaw". We recommend that you retain this filter to quickly locate target instances.​
    Relevant information​
    ​
    Associated project​
    You can add the application update task to the project and manage it in a centralized manner by project dimension, which helps maintain resource independence and ensure data security. Learn more.​
    tip​
    • After you select a project, new instances that meet the conditions automatically match the application update task.​
    • The project displayed here is the one you selected in Step 3. You can change the project in the top menu bar.​
    Tag​
    Optional. Used to select the scope of instances the application update task applies to based on tags. Click the "Add tag" button, and then select or enter the tag key and its corresponding tag value.​
    tip​
    • Up to 20 tags can be added.​
    • A tag is a key-value pair used to identify cloud resources. Tags help you categorize resources based on different dimensions, such as purpose, owner, or environment. This allows you to search and aggregate your resources. Learn more.​
    ​
  1. Click the "Confirm" button to complete the operation.​
  1. After you confirm that the information to be updated is correct, click the "Confirm" button to start deploying security protection.​
Expected results​
  • Security protection deployed​
  • The deployment takes about 3 minutes. When the task status shows Succeeded, the deployment is complete.​
  • Security protection deployment failed​
  • A failed deployment does not affect the normal operation of the original services on the instance. It only means security protection is not enabled this time. You can try deployment again or submit a ticket for technical support.​
Related operation​
You can also enable security protection when purchasing an ECS instance to obtain enterprise-level security protection for OpenClaw. For details, see ​
​
Last updated: 2026.06.02 15:52:32