From the course: Handling Sensitive Data with Cloud and Local AI

Unlock this course with a free trial

Join today to access over 25,500 courses taught by industry experts.

Visualizing LLM risks: Create an interactive UI

Visualizing LLM risks: Create an interactive UI

From the course: Handling Sensitive Data with Cloud and Local AI

Start my 1-month free trial Buy for my team

Visualizing LLM risks: Create an interactive UI

“

Renowned developer Simon Wilson, who shows us this lethal trifecta of LLM risks, also has a blog post that shows us how we can design systems that help prevent prompt injections. Now, I encourage you to read this blog post diligently. I want to show you something cool that we can do to kind of visualize And I like to do this for this pattern. eye that illustrates the following pattern. And I really like creating diagrams and interactive little programs in order to visualize some of these design patterns. So here we have the user asks for a summary of a malicious website. We have that the Privileged Agent coordinates the request. It never sees the raw website content. It stores it in a variable. The Quarantined LLM reads the problematic content and produces a summary as a variable. So basically, we are guarding this Privileged LLM from the malicious content itself. Then the Privileged LLM only sees VAR1. look at this in code soon and instructs the UI to show it. And finally, we have a…

Contents