On March 10, the National Internet Emergency Center released a risk advisory titled "Risk Alert Regarding OpenClaw Security Applications." The notice highlights that OpenClaw, also known as "Little Lobster," has recently gained significant popularity in downloads and usage, with major domestic cloud platforms offering one-click deployment services. This intelligent agent software operates computers directly based on natural language instructions. To enable autonomous task execution, the application is granted high system permissions, including access to local file systems, reading environment variables, calling external service APIs, and installing extensions. However, due to its extremely weak default security configuration, attackers can easily gain full system control once a vulnerability is exploited.

Several serious security risks have already emerged due to improper installation and use of the OpenClaw agent: 1. Prompt Injection Risk: Attackers can embed hidden malicious commands in web pages. If OpenClaw is induced to read such a page, it could be tricked into leaking user system keys. 2. Misoperation Risk: Due to potential misinterpretation of user commands and intent, OpenClaw might permanently delete critical information such as emails and core production data. 3. Plugin Poisoning Risk: Multiple plugins designed for OpenClaw have been confirmed as malicious or pose potential security threats. Installation could lead to key theft, deployment of trojan backdoors, and turning devices into "botnets." 4. Security Vulnerability Risk: To date, OpenClaw has publicly disclosed multiple medium and high-severity vulnerabilities. If exploited by attackers, these could lead to system takeover and leakage of private or sensitive data.

For individual users, this could result in the theft of private data, payment account details, and API keys. For critical sectors like finance and energy, consequences could include exposure of core business data, trade secrets, and code repositories, potentially paralyzing entire business systems and causing incalculable losses.

The advisory recommends that organizations and individual users implement the following security measures when deploying OpenClaw: 1. Strengthen network controls; avoid exposing OpenClaw's default management port directly to the public internet. Implement security measures like authentication and access control. Use strict environment isolation, such as containers, to limit excessive permissions. 2. Enhance credential management; avoid storing keys in plaintext within environment variables. Establish comprehensive operation log auditing mechanisms. 3. Strictly manage plugin sources; disable automatic updates. Only install signed extensions from trusted channels. 4. Continuously monitor for patches and security updates; promptly perform version upgrades and install security patches.

An expert from the Ministry of Industry and Information Technology (MIIT) recently emphasized the need for cautious use of AI agents like "Lobster." While the open-source AI agent has garnered significant attention and practical application within China's industry and user base, fostering the growth of the domestic AI agent ecosystem, its powerful execution capabilities also present serious security challenges.

The expert clarified that updating "Lobster" to the latest official version can fix known vulnerabilities but does not eliminate all risks. As a locally-run AI agent with autonomous decision-making and system resource access, coupled with模糊的信任边界 and a largely unvetted skills market, significant risks remain. Examples include potential misinterpretation of user instructions leading to harmful actions like deletion, using skill packs containing malicious code leading to data breaches, and risks from misconfigurations such as internet exposure or storing keys in plaintext—even with the latest version.

Cybersecurity is dynamic, and hacker techniques constantly evolve. Patching and upgrading cannot be seen as a one-size-fits-all security solution. Authorities urge government agencies, enterprises, institutions, and individual users to exercise caution when using agents like "Lobster." Discovered vulnerabilities or security incidents should be reported immediately to the MIIT's cybersecurity threat and vulnerability information sharing platform for timely handling according to regulations.

For safe usage, besides timely updates, adhering to the principles of "least privilege, proactive defense, and continuous auditing" is crucial. Key recommendations include: - Using the latest official version from official channels, enabling update notifications, backing up data before upgrades, and restarting services afterward. - Strictly controlling internet exposure; never expose the agent instance directly to the public internet. Use SSH or VPN for necessary access, restrict source addresses, and employ strong authentication. - Adhering to the principle of least privilege; avoid using administrator accounts, grant only necessary permissions, require secondary confirmation for critical operations, and consider isolation via containers or virtual machines. - Exercising caution in skill markets like ClawHub; scrutinize skill pack code before installation and reject any requesting downloads, shell script execution, or password input. - Establishing long-term protection mechanisms; enable detailed logging, conduct regular vulnerability checks, and utilize security tools and antivirus software. Regularly monitor official security bulletins and MIIT vulnerability databases for warnings.

Users are strongly advised to keep security as their bottom line, thoroughly understand and implement security configuration requirements, and develop safe usage habits. Ongoing security monitoring will continue, with timely warnings and technical support provided for any emerging risks.