In a transaction, W&I insurance (warranties & indemnities, or RWI in the United States) exists to transfer to an insurer the risk that the seller's representations turn out to be false. It's the lubricant that lets many deals close. And that lubricant is running thin on one specific point: artificial intelligence.
A policy exclusion is a risk the insurer refuses to carry. It doesn't vanish for all that — it stays on the table, and someone has to absorb it: the buyer through a discount, or a heavier escrow. On AI representations — training-data provenance, model performance, regulatory compliance — that's exactly what's taking shape.
In concrete terms: an AI asset whose compliance is neither documented nor attested becomes partially uninsurable. The insurer excludes what it can't assess. And what can't be insured gets priced as a markdown.
A W&I underwriter can only cover what it can characterize. A documented D7™ score — AI Act exposure across 7 dimensions, evidence levels PROVEN/INFERRED/ABSENT — gives it the factual basis that turns a vague "AI representation" into an assessable risk. That's the difference between an exclusion and coverage.
For the seller, the stake is direct: arriving with a scored and attested asset means making the representation insurable — and taking away the buyer's argument for a discount on uncovered risk.