Every AI data deal makes bets across three rings: what the contract permits (Rights), how the system is built (Architecture), and who captures the value (Commerce). Conventional legal analysis stays in the Rights ring. The structural outcome is determined in the other two.
The Termination Trap and RAG both start with conditional permission. One choice — training vs inference — determines everything that follows. The expanding gap is the structural risk that the contract does not address.
Source: AI Architecture Wheel diagnostic framework
Each deal archetype scored by the structural risk its positions create for the data supplier. Higher scores mean less control, less reversibility, less value captured. The market clusters to the right.
Source: AI Architecture Wheel — 12 named deal archetypes
Risk score: average of each archetype's positions mapped to a 0–100 supplier-risk scale. Training, Transformed, Entangled, Opaque, and Retained score highest. Clean, Inference, Observable, and Returned score lowest.
Across all 12 deal archetypes, Entangled, Retained, and Project appear most often. The market's structural defaults are lock-in, processor captures all value, and large capital commitments. The positions that favour the data supplier — Clean, Returned, Observable — barely appear.
Source: AI Architecture Wheel — position frequency across 12 archetypes