AI liability ILS
The far end of the pathway: insurance-linked securities — capital-market transfer of AI systemic risk. Read this page with its framing intact: the draft brochure scopes this as a 5–10 year horizon and a theoretical working paper, not a product on anyone’s shelf. GROUNDED · Brochure
The Hurricane Andrew arc
Section titled “The Hurricane Andrew arc”The template comes from natural catastrophe. Hurricane Andrew (1992) produced losses far beyond what the market had modeled — and the aftermath forced the industry to build catastrophe-modelling infrastructure it had never needed before. Once credible cat models existed, catastrophe bonds followed: instruments that let institutional capital (pension funds, asset managers) hold slices of nat-cat risk, priced off the model rather than off an underwriter’s intuition. STABLE
The claimed parallel: the accumulation problem — one foundation-model update changing every downstream deployment simultaneously — is AI’s systemic, correlated peril. It is too large and too correlated to sit comfortably on individual syndicate balance sheets, which is exactly the shape of risk capital markets exist to absorb. But capital-market transfer has a hard prerequisite: a loss model someone can price against. GROUNDED · Brochure
What the structure would be
Section titled “What the structure would be”| Element | The draft’s version |
|---|---|
| Data layer | The OVERT-attested inference population — a verified denominator across an entire book. GROUNDED · Brochure |
| Model | A loss model built on that population as claims experience accumulates (see claims-verified indemnity). |
| Instrument | ILS transferring AI systemic risk to institutional capital, “the way catastrophe bonds connected it to natural-catastrophe risk.” GROUNDED · Brochure |
| Capacity effect | Volume 1’s reinsurance & capacity chapter, one level up: capital markets as the ultimate reinsurer. |
The draft’s own words for why this belongs in an Experimental Innovation track: Inference Re “is not a new product for an existing line. It is a new asset class: AI systemic risk made transferable to institutional capital for the first time, through a verified inference population.” GROUNDED · Brochure Note the hedge discipline: that is a thesis about what the evidence infrastructure enables, and it should always be presented as one — see the claim taxonomy.
How to talk about this without overclaiming
Section titled “How to talk about this without overclaiming”- Do say: the ILS structure is explicitly scoped as a five-to-ten-year horizon and an open theoretical working paper. Skeptics who call it distant are agreeing with the brochure, not refuting it.
- Do say: the prerequisite chain is concrete — parametric triggers now, verified loss ratio as claims accumulate, loss model after that, capital-market transfer last. Each link is checkable.
- Don’t imply any ILS instrument exists, is planned for a date, or has investor interest. None of that is assertable. VERIFY before repeating anything more specific than the arc itself.
Drill this page →17 bank questions stand behind what you just read. Check it while it’s warm.