Skip to content

AI Security

LLM security: prompt injection, jailbreak detection, guardrails, and adversarial evaluation.

Subscribe
All Breaking Security Features
Important Recent
← Releases
Review required
Acacian/aegis v0.9.1 Security relevant
Auth RBAC

Privilege escalation + resource gaming fixes

Open
AI-Infra-Guard by Tencent Zhuque Lab v4.1.1 Security relevant
Security fixes
  • CVE-2026-XXXXX – Mask token fields in GetTaskDetail response to prevent credential leakage (#226)
Notable features
  • New Vulnerability Rules batch expanding AI component detection coverage
  • Fingerprint Enhancement with correct new-api matcher syntax (FOFA 100%)
v0.74.1 (2mo) Security hardening
v0.71.3 (2mo) OS CVE patches + attestations
v0.70.4 (2mo) Vuln DB hardening
v0.66.0 (2mo) Security hardening
v0.59.3 (2mo) Audit & security fixes
v0.59.0 (2mo) Security hardening
v0.58.1 (2mo) ClawHub trust hardening
v0.31.3 (3mo) ClawHub trust hardening
AI-Infra-Guard by Tencent Zhuque Lab v3.5-preview-2 Security relevant
Security fixes
  • CVE-2025-55182 — detection added for React2Shell vulnerability affecting Dify, NextChat, and LobeChat
v3.4.1 (8mo) CVE-2025-23316 fingerprint
v3.2 (9mo) AIG IP check fix
beelzebub v3.4.0 Security relevant
Security fixes
  • Added decoy MCP tools to detect and collect prompt injection attacks
Notable features
  • MCP honeypot tools for injection detection
  • Real-time attack prompt collection

Beta — feedback welcome: [email protected]