# V2 AI Model Cards

Every AI service used by V2 gets one model card in this directory before live
use. Use the fields required by `V2/docs/decisions/model-card-hosting.md`.

Current cards:

- `adaptive-ai-director.md` — V2 Adaptive AI Director limited-risk disclosure,
  opt-out, `@nous/safety` hosting metadata, and `@themis/accountability`
  AI-system-of-record evidence.
- `ai-commentary.md` — V2 AI commentary disclosure, opt-out/fallback,
  off-rollback presentation boundary, `@nous/safety` hosting metadata, and
  `@themis/accountability` conformity evidence.
- `anti-cheat-classifier-suite.md` — V2 smurf, win-trading, coordinated-throw,
  and geographic-anomaly fair-play classifiers, with `@nous/training` training
  ownership, `@nous/safety` hosting metadata, and `@themis/accountability`
  conformity evidence.
- `generation-pipelines.md` — V2 AI-assisted cosmetic, stage/decal, and
  commentary lipsync generation pipeline disclosure, provenance fallback, shield
  gates, `@nous/safety` hosting metadata, and `@themis/accountability`
  conformity evidence.