In Depth
Insurance runs on loss history. A carrier can price auto or workers' comp because it sits on decades of data about how often a given exposure turns into a claim and how big that claim gets. AI liability has almost none of that history. The failure modes are new, the case law is thin, and the few decisions that exist are scattered across courts, tribunals, and regulators in different jurisdictions. Underwrite without that, and you're guessing.
Stria is Ollive's answer to that gap. It collects the AI cases as they emerge: a chatbot's fabricated promise, a biased screening model, a model that disclosed data it shouldn't have, a regulator's enforcement action. Each is recorded in a structured form: the type of agent, the alleged failure, the theory of liability, the posture, and the outcome where one exists. Over time that turns a pile of headlines into something you can query and reason about.
That intelligence flows into two places. It informs underwriting, so the price of a policy reflects how comparable agents have actually fared in court rather than a flat assumption. It also feeds the Agent Trust Score: if a particular pattern of agent behavior keeps showing up on the wrong end of a claim, agents that share that pattern should score lower and price higher.
What It Looks Like
A vendor brings Ollive an AI agent that drafts customer-facing answers about a regulated product. To price it, the relevant question is concrete: have agents that make confident, customer-facing statements been held to those statements? Moffatt v. Air Canada (2024 BCCRT 149) is one verified data point: a tribunal held the airline to its support chatbot's invented policy. A litigation-intelligence database exists to gather that kind of resolved case systematically, so the next quote rests on a pattern of outcomes rather than a single anecdote or a hunch.
As the body of AI case law grows, the same lookup gets sharper. The point of Stria is that the hundredth AI claim should be priced with more knowledge than the first.
Why It Matters For AI Vendors
When you buy AI coverage, you want it priced by someone who understands how AI claims actually play out, not by a carrier extrapolating from unrelated software losses and padding the premium to cover its own uncertainty. Litigation intelligence cuts that uncertainty premium. The better the loss picture, the more a low-risk agent gets rewarded for being low-risk instead of paying for the whole category's worst case.
It also tells you where your own exposure sits. Knowing which AI failures are drawing claims, and under which legal theories, is useful long before you ever file one. It tells you what to test for, what to disclose to buyers, and where to put a human in the loop.