Risk Factors in Mend AI

Overview

Risk Factors in Mend AI help you prioritize findings by highlighting models in your inventory that may warrant a more immediate call-to-action compared to other models.

They are listed in multiple locations across the Mend AI Native AppSec Platform user interface:

1. The Risk Factors column in the AI Models table:

2. The model’s side-panel:

   

Mend AI Risk Factors

• No Findings – No known vulnerabilities detected.

• False-Positive – Safe, i.e., reported by Hugging Face as unsafe but refuted by the Mend AI Research team.

  

• Confirmed Unsafe – Unsafe, i.e., reproduced and verified by the Mend AI Research team.

  

• Unconfirmed Unsafe – Suspected Unsafe, i.e., tagged by Hugging Face as unsafe, but not reviewed by the Mend AI Research team.

  

• Conversational System Prompt - System prompt classified as a conversational AI interface. These interfaces tend to have higher risk exposure due to their fuzzy and open-ended nature.

  

AI Risk Factors in Automation Workflows

The Mend AI Risk Factors can be used in the Mend AppSec Platform’s Automation Workflows, by selecting Security → AI Analysis as the Triggering Event.

In the example below, the event condition is the detection of a conversational interface (system prompt) or a confirmed unsafe Hugging Face model.