Is my AI tool safe?Assistants and personal agentsOpen source (formerly Clawdbot and Moltbot)
Is OpenClaw safe?
No tool is safe or unsafe on its own: it depends on what it can reach and how it is set up. The record for OpenClaw: 3 incidents in this center's library and 1 CVE in its advisory feed (1 high), the newest published Feb 2, 2026. Below: what went wrong, what was fixed, and how to use it safely.
What to watch for
The most-starred AI agent on GitHub: a personal assistant wired to shell, browser, files and every chat app. One-click takeover, hundreds of malicious skills in its marketplace, and the agent of choice in the Taiwan government intrusion.
How to use it safely
Run it in a VM with only the integrations you need, install skills only after reading them, and keep it away from accounts that hold money or production access.
What it can reach
- Runs commands with your permissions
- Connected to mail, files, chat or CRM
- Installs plugins, skills or tools from registries
- Holds API keys and tokens
What went wrong: incidents in the library
- Taiwan: a team of AI agents breaks into government systems in four days Critical
Taiwan confirmed that a July 2026 hacking campaign against its government combined human operators with a team of AI agents, including the OpenClaw agent. Over about four days the attackers stole passwords and personnel records,… - OpenClaw: one-click takeover of a personal AI agent High
OpenClaw, a popular self-hosted personal AI agent that reads messages, runs commands and holds API keys, could be taken over with a single malicious link. Researchers also found tens of thousands of OpenClaw instances exposed on… - ClawHavoc: hundreds of malicious agent skills deliver infostealers High
Attackers flooded ClawHub, the skill marketplace for the OpenClaw agent, with hundreds of fake skills posing as crypto and productivity tools. Each told users to install a "required" helper that was actually password- and…
Known CVEs
- CVE-2026-25253 High OpenClaw (Clawdbot) one-click RCE via gateway token exfiltration
How many people use it
- 391k GitHub stars · GitHub
A high CVE count can mean heavy scrutiny as much as weak engineering, so read the incidents and advisories, not only the number. The counts are what this center has recorded, not every flaw ever found; usage figures are the vendors' and registries' own published numbers.
Sources
- GitHub: GitHub stars
- GitHub Advisory Database: clawdbot
- Al-Monitor: Taiwan says it was targeted by an AI-driven hacking campaign
- The Next Web: Taiwan says AI agents helped hack its government in four days
- OpenClaw: advisory GHSA-g8p2-7wf7-98mq
- NVD: CVE-2026-25253
- depthfirst: one-click RCE to steal your OpenClaw data and keys
- The Hacker News: 341 malicious ClawHub skills
- OpenSourceMalware: malicious ClawHub skills target OpenClaw users
- The Hacker News: OpenClaw adds VirusTotal scanning