MLSecOps and AI security, built in the open

Everything MLSecOpsHub builds lives in public repositories on GitHub: the awesome-MLSecOps list of open-source MLSecOps and AI security tools, the Agentic Attack Index of documented AI agent attacks with its Rogue Agent Watch dashboard, and vulnerable-by-design labs for practising AI agent security. Star a project to follow it, or open an issue to contribute.

• awesome-MLSecOps

  A curated list of MLSecOps tools and resources for securing machine learning and AI systems - adversarial ML defense, LLM security, AI red teaming, model scanning, supply-chain protection, and MLOps pipeline security.

  Astro 480 ★ 123 forks Updated Oct 19, 2026 Website

• mlsecopshub

  MLSecOpsHub website

  • ai-security
  • mlsecops
  • open-source

  Astro Updated Oct 23, 2026 Website

• agentic-attack-index

  Open, source-linked dataset of real-world cyberattacks executed or orchestrated by AI agents. Every field cited, every grade honest. Data layer of Rogue Agent Watch.

  • agentic-ai
  • ai-agents
  • ai-security
  • autonomous-agent-identity
  • autonomous-agent-security
  • autonomous-agents

  JavaScript Updated Oct 22, 2026 Website

• rogue-agents-watch

  Rogue Agent Watch: tracking down rogue agents incidents, autonomous AI attacks and AI agents sandbox evasion

  • agent-identity
  • agent-security
  • ai-governance
  • ai-security-research
  • ai-security-tool
  • autonomous-agent-identity

  TypeScript 1 ★ Updated Oct 22, 2026 Website

• provenance-engineering

  Open knowledge base for Provenance Engineering: data lineage, identity governance, model provenance and software inventory.

  • agent-provenance
  • aibom
  • cbom
  • data-lineage
  • identity
  • identity-governance

  1 ★ Updated Oct 16, 2026

• world-models-security

  Early-stage repository; no description published yet.

  Updated Oct 15, 2026

• agentic-access-control

  Read-only, local-first dashboard showing what your AI agents (Claude Code, Gemini CLI, MCP servers) are actually allowed to do: permission rules with file provenance, MCP server inventory, sandbox & hook posture, risk findings, OWASP Agentic Top 10 threat scenarios, and config drift. No enforcement, no telemetry (for now).

  • access-control
  • agentic-access
  • ai-agents
  • ai-agents-security
  • ai-observability
  • ai-threat

  TypeScript Updated Oct 11, 2026

• damn-vulnerable-skills

  Vulnerable-by-design security lab for Claude Agent Skills (SKILL.md) and Agent Skills in general

  • agent-security
  • agent-skills
  • ai-security
  • claude-code
  • claude-skills
  • llm-security

  Python Updated Sep 11, 2026

Definition

What is MLSecOps?

MLSecOps is the practice of applying security engineering to the machine learning lifecycle: training data, model artifacts, pipelines, deployment, and the LLM applications and agents built on top. It extends DevSecOps with AI-specific threats such as model supply-chain attacks through unsafe serialization, prompt injection, and over-privileged agents and MCP tools.

Read the reference explainer, What is MLSecOps?, on The MLSecOps Hacker, and browse tools by category on awesomemlsecops.com.