Adversarial Design Thinking
A methodology for
AI red teaming
Free, methodology-driven education for AI red teaming. Learn the techniques, the craft, and how to think about targets.
New here?
What AI red teaming is, why it matters, and the core concepts you need before starting.
Explore
Framework
The ADT methodology. Five phases of adversarial thinking applied to AI systems.
Prompting
How prompts work and why it matters. Every concept has a constructive face and an adversarial one.
Techniques
Attack categories and methods. Encoding, persona, framing, multi-turn exploitation.
Crafting
How to construct adversarial prompts. The craft behind effective red teaming.
Jailbreaks
System-level prompt construction. How jailbreaks work and how to build them.
Exercises
Hands-on practice. Apply the ADT methodology to real AI systems.
Compose
Build adversarial prompts layer by layer. Browse techniques, compose meta-prompts, generate structured instructions.
Field Notes
Practitioner dispatches. CTF walkthroughs, jailbreak experiments, methodology in action.
Recent Field Notes
The Attacker Moves Second
Fourteen researchers broke every AI defense they tested. The attack process they used has been practiced in other fields for decades.
Selara CTF: ADT Applied to Defense Escalation
Walking through Redline's Selara CTF using adversarial design thinking. Eight levels, eight defense mechanisms, zero existing walkthroughs.
Stay current
New field notes weekly. AI red teaming techniques, practitioner insights, and methodology updates.
Responsible use: These techniques are documented for defensive understanding and authorized security testing. Only test systems you own or have explicit permission to test. Read the full statement.