LLM red teaming & AI security · 2026 edition
It will be fooled. Build the system around it so that when that happens, nothing important breaks. That line opens OWASP's 2026 Top 10 for LLM Applications.
Every defense in this course carries a label: does it still hold once the attacker knows it is there?
10 tracks · 93 modules · ~28h
Two modules, mostly hands-on. What a language model actually does, what your app really sends it, and what changes when it thinks first.
2/2 written
The conceptual spine. Eight models that make every later module make sense.
8/8 written
Attacks on the model's behaviour: jailbreaks, encoding, multi-turn, reasoning-model surfaces.
10/10 written
The core of the course. Where the attacker never talks to the agent, and where the money is.
12/12 written
The attacks that never go through the prompt: poisoning, backdoors, artifacts, extraction.
0/7 written
The controls that actually survive an adaptive attacker, plus an honest table of the ones that don't.
0/14 written
How the work is actually run: threat models, invariants, adaptive testing, measurement, reporting.
0/12 written
Tools chosen by the question they answer, and why generic scanners miss your real bug.
0/13 written
What attackers are actually doing with AI right now. The fastest-decaying content here, refreshed monthly.
0/5 written
Regulation as it actually stands, programme design, and how to get hired doing this.
0/10 written