A surge of interest in the open-source AI agent tool OpenClaw has raised growing concerns over privacy and cybersecurity risks, prompting authorities to issue warnings about potential vulnerabilities.
Online users have nicknamed the tool "lobster" because its icon resembles a red lobster. The agent can autonomously handle tasks such as file management, email processing and data analysis.
However, the trend has also raised concerns. On Wednesday, the topic "the first batch of lobster users has begun uninstalling" trended online after some users reported issues such as accidental email deletion and possible privacy leaks.
The tool's sudden popularity has also spawned new services on second-hand trading platforms, where technicians offer "on-site installation". "On-site uninstallation" services have also become popular as users grow more cautious.
Authorities have also issued risk alerts. On Feb 5, a platform under China's Ministry of Industry and Information Technology warned that some instances may face cyberattack and data leak risks due to weak configurations.
On Tuesday, the National Computer Network Emergency Response Technical Team/Coordination Center of China (CNCERT/CC) said weak default settings could allow attackers to gain full system control, citing risks such as prompt injection, unintended operations, malicious plugins and software vulnerabilities.
Experts have suggested several measures to reduce potential threats when using such AI agents.
Users are advised to download the latest version from official sources, enable updates and back up data before upgrading. They should also avoid exposing instances to the public internet and use strong authentication methods.
Experts recommend following the principle of least privilege by granting only the permissions necessary for tasks and requiring confirmation for sensitive operations.
They also warn that skill packages on community platforms may contain malicious code, urging users to review them carefully before installation.
In addition, users should avoid suspicious websites and links, enable security tools and regularly check for vulnerabilities.