OpenClaw: Why Small Businesses Should Be Very Careful
The Reality of AI Agents
Personal AI agents like OpenClaw are being promoted as smart helpers that can read your files, run commands and plug into other systems to do work for you.
However, recent work by major security vendors paints a different picture: OpenClaw, as it exists today, is high risk for normal business use on everyday laptops and desktops. It should be treated as untrusted software with broad permissions, not a harmless productivity app.
What OpenClaw Actually Does
OpenClaw is more than a chatbot. Once connected to your machine, it has the power to:
- Read and write local files
- Run scripts and system commands
- Call "skills" that talk to other apps and services
- Remember information across sessions
Essentially, you are granting it the same power as a human with a logged-in account on your device.
The Main Risks in Plain English
1. It can be tricked into harmful actions
A malicious web page or a carefully crafted prompt could convince OpenClaw to run dangerous commands, download untrusted code or delete files—all while "just following instructions".
2. It can leak passwords and secrets
If OpenClaw can see your files or browser data, it can also see saved passwords and API keys. These can be used to log in to your email, CRM or accounting tools without your knowledge.
3. Add-on "skills" are weak points
Third-party skills may be poorly written or intentionally malicious. Enabling a bad skill is effectively installing hostile code inside a tool that already has wide-ranging access.
Practical Guidance for Safe Testing
- Isolate the environment: Never run OpenClaw on your primary work machine; use a separate test device or virtual machine.
- Limit permissions: Use low-privilege accounts and do not connect it to banking, payroll or admin portals.
- Audit your skills: Only enable skills you have checked and actually need. Turn off anything you do not understand.
Conclusion
For many small businesses, the honest answer is you should probably not use OpenClaw at all - at least not with live data. If you are curious about the potential value, keep it away from real client or financial data and run it in a tightly controlled test environment.
Identify Your AI Risks
Are you worried about how AI agents might be exposing your business? SME Cyber Solutions provides practical security reviews to ensure your digital transformation is safe.
Request a Security Review