A Comparative Evaluation of AI Agent Security Guardrails

TL;DR

This paper compares the effectiveness of DKnownAI Guard with other security guardrails for AI agents, showing it outperforms competitors in detecting threats and harmful requests.

Contribution

It provides a comprehensive benchmark evaluating multiple AI agent guardrails against human-annotated ground truth, highlighting DKnownAI Guard's superior performance.

Findings

DKnownAI Guard achieves 96.5% recall in risk detection.

It ranks first in true negative rate at 90.4%.

Overall, it demonstrates the best performance among evaluated guardrails.

Abstract

This report presents a comparative evaluation of DKnownAI Guard in AI agent security scenarios, benchmarked against three competing products: AWS Bedrock Guardrails, Azure Content Safety, and Lakera Guard. Using human annotation as the ground truth, we assess each guardrail's ability to detect two categories of risks: threats to the agent itself (e.g., instruction override, indirect injection, tool abuse) and requests intended to elicit harmful content (e.g., hate speech, pornography, violence). Evaluation results demonstrate that DKnownAI Guard achieves the highest recall rate at 96.5\% and ranks first in true negative rate (TNR) at 90.4\%, delivering the best overall performance among all evaluated guardrails.