Cortex Security Releases 'LLM-Guard v2.0', an Open-Source Firewall for Enterprise AI Applications

Overview

Severity: MEDIUM | Affected: Cortex Security | Category: tool

Cortex Security has announced the release of LLM-Guard v2.0, a major update to its open-source firewall designed to protect enterprise applications built on large language models. This new version introduces several key features, including a sophisticated PII detection and redaction module that supports over 50 entity types, advanced topic-based moderation to prevent policy violations, and a novel "semantic jailbreak" detection system that analyzes user intent rather than relying solely on keyword matching. LLM-Guard v2.0 acts as a programmable layer between users and the LLM, inspecting both prompts and model responses in real-time. It aims to provide developers with a robust, easy-to-integrate solution for enforcing security and safety policies, thereby reducing the risk of prompt injection, data leakage, and misuse of AI systems. The framework is available on GitHub and supports major LLM providers.

References

• https://github.com/cortex-security/llm-guard/releases/tag/v2.0.0
• https://cortex-security.com/blog/introducing-llm-guard-v2/