{
  "name": "African AI Trust Index — Methodology",
  "edition": "preview-2026-08",
  "claimTier": "T1_preview",
  "url": "https://benchmark.mlatho.com/methodology.html",
  "tracks": [
    { "code": "NY", "name": "Chichewa comprehension / QA", "targetCount": 35, "mapsTo": ["africanLanguage"] },
    { "code": "TR", "name": "Chichewa ↔ EN translation", "targetCount": 20, "mapsTo": ["africanLanguage"] },
    { "code": "CX", "name": "Cultural / civic context (EN)", "targetCount": 25, "mapsTo": ["culturalUnderstanding", "businessReadiness"] },
    { "code": "SF", "name": "Safety & refusal (Africa-relevant)", "targetCount": 20, "mapsTo": ["safety"] }
  ],
  "scoringModes": ["exactish", "rubric_1_4", "safety_class"],
  "composite": "0.35*Language + 0.25*Culture + 0.20*Business + 0.20*Safety",
  "runProtocol": {
    "modelsPinned": 4,
    "temperature": 0,
    "toolsAllowed": false,
    "browsingAllowed": false
  },
  "forbiddenClaims": [
    "Africa's first AI benchmark",
    "Statistically significant proof across the industry",
    "African LLM / foundation model launch",
    "Supports all African languages",
    "Model X is racist"
  ],
  "hedge": "The African AI Trust Index Preview is a small human-evaluated set focused on Chichewa and African context. Scores are for this set only, under a fixed prompt template without tools. They are not a universal ranking of model quality."
}