Learn AI security together.
A space for learning AI and cybersecurity together — sharing what we build, what we break, and what we find along the way. Whether you are just getting started or already shipping AI features, there is something here for you.
What we teach
Every topic maps directly to a hop in the AI request path. We deconstruct real vulnerabilities and examine defenses at each stage.
Prompt Injection
Hop 1How attackers manipulate prompts to override instructions, extract data, or change model behavior.
Retrieval Poisoning
Hop 2How malicious content in document stores can influence RAG outputs and compromise retrieval integrity.
Agent Permissions
Hop 4How over-permissioned tool calls let agents take actions they were never meant to, and how to constrain them.
Evaluation Basics
Hop 3How to measure whether your AI system's security controls actually work — before and after deployment.
Ways to take part
Prompt Injection Workshop
Hands-on session covering direct and indirect prompt injection techniques, detection methods, and defense patterns.
Securing RAG Pipelines
A written guide covering document validation, embedding integrity, and retrieval-layer defenses for RAG systems.
Agent Permission Exercises
Open exercises where you audit agent tool-call configurations and practice applying least-privilege principles.
Community Discussion
An open channel for discussing AI security topics, sharing findings, and connecting with others working on these problems.
Want to learn or contribute?
Join our open learning track. Reach out to collaborate on exercises, suggest topics, or share your security research.
Join Community