Overview
Severity: MEDIUM | Affected: SentinelAI Foundation | Category: tool
The SentinelAI Foundation has announced the general availability of LLM-Guard v2.0, a major update to its open-source security toolkit for AI applications. LLM-Guard acts as a self-hostable firewall that inspects both prompts sent to and responses received from large language models. This new version introduces real-time, low-latency scanning to detect a wider range of threats, including complex prompt injection attacks, PII leakage, toxic content, and adversarial data designed to poison models. Key features include a pluggable architecture for custom detectors, a centralized dashboard for managing security policies across multiple applications, and native integration with popular frameworks like LangChain and LlamaIndex. By providing a robust, transparent security layer, LLM-Guard aims to empower developers to deploy AI features more securely and mitigate common risks associated with user-facing LLM interactions.