THE STARTING POINT
Three things to get right.
- Define the assets and trust boundaries before selecting tests.
- Choose observable adversary behaviors rather than vague threat labels.
- Capture both detection evidence and the gaps your test exposes.
TRY THIS IN THE LAB
A question worth testing.
Map a synthetic AI assistant’s retrieval, model, and tool layers. Pick one abuse case at each boundary and write the observable evidence you would expect.
Browse the lab notesLeave a useful trail.
Write down your starting assumptions, scope, expected behavior, observed behavior, and evidence. A successful test should explain what happened and which question to investigate next.