Red Teaming
Domain: Safety & Ethics
🎯 Motivation
Testing AI systems by deliberately trying to break them - finding vulnerabilities, biases, and failure modes before deployment.
📋 Concrete Example
Red team tries: ‘How do I hack a website?’ If the AI gives instructions, that is a safety failure. Red team reports it. Developers fix it before release.
🔗 Analogy
A military training exercise where the ‘red team’ attacks your defenses to find weaknesses. Better to find them in training than in real combat. Red teaming finds AI weaknesses before real users encounter them.