AI Agents Now Have Their Own Insurance Category. What It Means That Underwriters Are Pricing Agent Failure

A wave of specialist insurers, including Corgi, Armilla, Testudo, and HSB/Munich Re, launched liability products this year specifically covering AI agent failures and hallucinations. Corgi, a YC-backed insurer, offers quotes in under 10 minutes with same-day binding. Armilla, a Lloyd's Coverholder, reports policy limits up to $25 million covering AI performance and liability including hallucinations. Testudo writes a standalone Lloyd's-backed generative AI policy up to $10 million.

Why an insurance market is a genuinely useful signal

An underwriter's job is to price risk with real capital behind the number, not to have an opinion about AI safety in the abstract. When multiple specialist insurers are willing to write policies against agent failure, at meaningful limits, with underwriting fast enough to quote in minutes, that's an external, money-backed data point about how the market actually assesses the maturity and risk profile of autonomous agents right now, distinct from vendor marketing or internal debate. It's not proof agents are safe, insurers price risk they expect to sometimes pay out on. It's evidence the risk is now considered measurable and insurable, which is itself new.

What this means for your own deployment decisions

If your organization is debating whether an agent deployment is mature enough to trust with a consequential task, the existence and pricing structure of this insurance category is a useful external reference point. Coverage terms, exclusions, and premium levels across these products effectively encode a real market assessment of failure risk by task type and deployment scope. Worth looking at how these policies define and price agent risk even if you're not buying one, since it's a more grounded signal than most internal risk conversations have access to on their own.