A field guide to the AI security jungle

What it does This is a curated index of frameworks, tools, standards, and learning resources for AI security — from adversarial ML toolkits to governance checklists to deliberately vulnerable MCP servers for CTF practice. Think of it as a map to a territory that is expanding faster than most teams can track.

The interesting bit The list is aggressively current: it already includes dedicated sections for agentic AI and MCP (Model Context Protocol) security, plus Singapore’s government-backed AI Verify framework and OWASP’s nascent AI Vulnerability Scoring System. Someone is clearly paying attention to where the field is heading, not just where it has been.

Key highlights

• Covers the full lifecycle: governance frameworks (NIST, ISO, Google SAIF), red-teaming tools (garak, cleverhans, Counterfit), defensive controls (guardrails, sandboxing, supply-chain scanning), and incident response guides
• Includes hands-on labs: vulnerable MCP servers, LLM CTF challenges, and Microsoft’s AI Red Teaming playground infrastructure
• Surfaces taxonomies and standards that are still emerging, like MITRE ATLAS, the Arcanum prompt-injection taxonomy, and OWASP’s agent observability standard
• Maintained by a single curator with a Mastodon handle, which either signals dedication or a bus factor of one

Caveats

• No quality filtering beyond curation; some listed tools may be abandoned or experimental
• The README is a flat link dump with minimal annotation, so you will need to click through to assess relevance

Verdict Worth bookmarking if you are building or securing AI systems and need a single starting point for standards and tooling. Skip it if you want depth on any one topic — this is a table of contents, not a textbook.