The AI Red-Team Question Deck
Recall the core questions that structure an AI red-team session from scope to evidence.
Scope What can this AI system actually do? Name data access, tools, outputs, users, and downstream decisions. Model behavior matters, but authority comes from the whole workflow. Prompt vs path Are we testing a prompt trick or a workflow path? Use prompt tricks when they map to a real path. Otherwise they are weak evidence. Objection We already tested the base model, so why test the product? A stakeholder wants to skip end-to-end testing. Your line The base model did not have our retrieval, tools, data permissions, or approval workflow. The red-team question is whether those controls hold together. Do…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in