LLM-security categories (2026)
LLM Threat sorts the market into six categories aligned to the OWASP GenAI Solutions Reference Guide: AI Red Teaming, Guardrails / LLM Firewall, AI-SPM / Runtime Protection, AI Governance / Model Risk, Agentic AI Security, and Open-Source Tools. Each page lists that pillar's vendors and tools, scored, as of 2026-07-13.
AI Red Teaming
Tools and services that attack your model before a real adversary does, hunting injection paths, jailbreaks, and data leaks.
Guardrails / LLM Firewall
Request-time filters that sit between users and models, blocking injected prompts, unsafe output, and PII leaks before they land.
AI-SPM / Runtime Protection
Posture management with live detection: who's running which model where, and whether anything is attacking it right now.
AI Governance / Model Risk
Inventory, policy mapping, evidence, and audit workflows for GRC and model-risk teams — governance rather than guardrails.
Agentic AI Security
Controls for AI that acts — tool-calling agents and MCP traffic — where excessive agency turns a bad prompt into a bad action.
Open-Source Tools
Public GitHub projects for testing and defending LLMs, tracked with live stars, license, and repo status so a dead project can't sneak into your stack.