Skip to content
Sign inOpen the app
Artwork for The Technology Risk Podcast
The Technology Risk Podcast · February 28 · 18 min

Episode 66 - OpenClaw is a Massive Security Risk

OpenClaw is an agentic AI, meaning it can take actions on your behalf, including reading local files, browsing the web, and using stored credentials. On February 19 2026, Microsoft warned that it should not be run on a standard personal or enterprise workstation because of identity and runtime risks. In this episode, we explain what agentic AI is, why it changes the traditional risk model, and how tools that act autonomously can expand the blast radius of a single mistake or compromise. We define terms like malicious skill and prompt injection and break down how they can be used to manipulate automated agents. Make sure to subscribe to our podcast on Spotify and Apple Podcasts for more technology insights every Friday at 6 PM (PDT), as well as check out our website at www.frostyos.com.

0:00-18:12
Play in podnod

transcript

No transcript — this publisher did not publish one.

show notes

OpenClaw is an agentic AI, meaning it can take actions on your behalf, including reading local files, browsing the web, and using stored credentials. On February 19 2026, Microsoft warned that it should not be run on a standard personal or enterprise workstation because of identity and runtime risks. In this episode, we explain what agentic AI is, why it changes the traditional risk model, and how tools that act autonomously can expand the blast radius of a single mistake or compromise. We define terms like malicious skill and prompt injection and break down how they can be used to manipulate automated agents.

Make sure to subscribe to our podcast on Spotify and Apple Podcasts for more technology insights every Friday at 6 PM (PDT), as well as check out our website at www.frostyos.com.

more episodes

All episodes