AI Foundationspredict · compress · act

concepts → Mechanism

Mechanism

Red teaming

timeless1 connectiondraft

attacking the system on purpose to find how it breaks before others do

Where it sits

The prism has six jobs across and eight layers down. Its primary cell ismeasure × L6. Hatched cells cannot exist — a GPU does not learn, an institution does not infer.

Overlays: Security, privacy & robustness.

What must come first — and what it unlocks

Left to right is reading order, derived from the prerequisite_of edges. Nothing here is hand-ordered: the diagram is the graph.

nothing comes first — this is a starting pointRed teamingnothing depends on it yet — a leaf in the reading order

What kind of thing it is — and what it is made of

It is part of: Evaluation.

Where to read it

The chapter that introduces it, and any chapter that uses it again.

32How Do We Even Know?act 7 · The Alignment Layer

Where it comes from

paperRed Teaming Language Models with Language ModelsEthan Perez, Saffron Huang, Francis Song, Trevor Cai, Roman Ring · 2022

Every connection

All 1 edges touching this node, grouped by relation family — the sections above are highlights from this list. Colours match the relation families inthe atlas.

Structure · 1
is part ofEvaluationField

This page is a projection of one node in src/data/concepts.ts. It has no prose file of its own — 230 declared edges produce all 236 of these pages. Edit an edge and both endpoints change.