AI agents can now browse, call APIs, hold credentials, and spend money on your behalf — which makes them the most consequential new attack surface of this decade. This guide indexes every AI-agent security article on Hmmnm into one learning path: from “what is MCP” to delegated-identity governance and living-off-the-LLM tradecraft.
Start Here
- AI Agent Security: Why Your Autonomous Systems Are the New Attack Surface — why autonomous systems changed the threat model
- OWASP Top 10 for Agentic Applications 2026: Complete Security Guide — the reference risk taxonomy
- AI Agents Gone Rogue: The 2026 Threat Landscape Nobody Prepared For — real incidents, not theory
Protocol & Infrastructure Security (MCP, A2A)
- What Is MCP (Model Context Protocol)? — foundations first
- MCP Security and Pentesting: Threats, Test Cases, and Hardening — test cases you can run
- Software Supply Chain Security for AI Agents and MCP Servers — agent + MCP server supply chains
- Securing Multi-Agent Systems: A2A, MCP, Memory, and Cross-Agent Trust Boundaries — A2A, memory, cross-agent trust
Attack Techniques
- Prompt Injection Attacks Explained: How Attackers Hijack AI Applications — the canonical primer
- Prompt Injection in 2026: Real Attacks & Defense Strategies — how the attacks evolved
- RAG Security: How Attackers Poison Your AI's Knowledge Base — poisoning retrieval corpora
- Living Off the LLM: How Attackers Weaponize AI Infrastructure in 2026 — LLM infra as the attack platform
- AI Model Extraction and Distillation Attacks: How Your Model Gets Stolen — stealing the model itself
- AI Supply Chain Attacks: When Your AI Model Becomes the Backdoor — backdoored models
Defense & Governance
- AI Agent Identity and Least Privilege: The Three-Layer Model — the three-layer identity model
- The AI Agent Authority Gap: Why Your IAM System Is Not Ready — closing the IAM delegation gap
- Zero Trust Architecture for AI Systems: Implementation Guide — zero trust applied to AI
- Securing AI Agents in the Enterprise SOC — agents inside the SOC
- Red Teaming LLM Applications: A Practical Playbook (2026) — a practical red-team plan
Frameworks & Ecosystem
- Building Cybersecurity AI Agents with CAI: The Open-Source Framework Powering Bug Bounties and CTFs — open-source security agents
- DefenseClaw and the AI Agent Security Arms Race: RSAC 2026 Takeaways — industry direction from RSAC 2026
- Agentic AI Security: Attack Surface in Autonomous Systems — the wider autonomous-systems view
- TurboQuant Explained: Google AI Memory Breakthrough — memory tech that changes agent design
New analyses are added continuously — bookmark this guide as your index for AI agent security in 2026.