Topics
The eight areas this publication covers. Deliberately short: a taxonomy that grows without discipline stops being navigation and becomes a tag cloud.
Prompt Injection & Jailbreaks
Direct and indirect prompt injection, jailbreak techniques, and the defences that actually hold up under adversarial pressure.
LLM Application Security
Securing applications built on language models: the OWASP LLM Top 10, output handling, sandboxing, and trust boundaries.
Agentic AI & MCP Security
Tool abuse, excessive agency, confused-deputy attacks, and securing Model Context Protocol servers and agent runtimes.
Model Supply Chain
Data and model poisoning, backdoored weights, unsafe serialisation formats, and provenance for models and datasets.
AI Red Teaming
Methodology, tooling, and reporting for adversarial testing of AI systems, from scoping through to remediation.
AI for Defense
Using AI on the defensive side: detection engineering, SOC automation, triage, and where it demonstrably does not work.
AI Governance & Compliance
The EU AI Act, NIST AI RMF, ISO 42001, and translating regulatory obligation into engineering controls.
Adversarial ML & Privacy
Evasion and extraction attacks, membership inference, training-data leakage, and privacy-preserving machine learning.