aiuc vs armilla vs testudo vs corgi:who covers what

The AI insurance category is young enough that buyers routinely compare companies that are not actually competitors. AIUC, Armilla, Testudo, and Corgi get named in the same breath, but they occupy different structural positions. Here is a plain map. We write cover in this category too, so read this knowing where we sit, and note that details move fast in this market.

aiuc: the standard-setter

A San Francisco managing general agent that bundles a standard, an audit, and insurance. Its AIUC-1 certification is the most talked-about attempt at a “SOC 2 for AI agents,” spanning six risk pillars, with product liability limits up to 50 million dollars on specialist paper. Its distinguishing move is the certification: the audit that sets your price is also the trust asset you show enterprise buyers. Its structural feature, and the thing to weigh, is that the same firm sets the standard, runs the audit, and carries the risk.

armilla: the verifier with a warranty

A Lloyd's coverholder and one of the few managing general agents focused exclusively on AI. It pairs affirmative AI liability cover with independent model verification, reports limits up to 25 million dollars, and is backed by established capacity including Chaucer, Axis Capital, and Convex. It also offers a performance warranty product, letting AI vendors put contractual guarantees in writing. Like AIUC, it verifies the models it insures.

testudo: the litigation-data underwriter

A Lloyd's coverholder writing standalone generative AI liability on an excess and surplus basis, backed by a syndicate panel, with capacity that has grown past 9 million dollars per insured. Its distinguishing move is data: it prices from a large dataset of real AI litigation and incidents through its own risk engine, and it deliberately does not require deep integration with your systems, which makes it fast to adopt.

corgi: the full-stack carrier

The odd one out, and the only true full-stack carrier in this group. Corgi is a licensed insurance carrier for startups that issues on its own paper, offering directors and officers, cyber, general liability, and AI liability in one fast, founder-friendly flow. Its AI cover is one shelf product in a broad startup stack rather than a specialty focus, so the depth of agent-specific underwriting differs from the pure-play specialists.

how to read the field

A few honest patterns. The limits race runs roughly from 50 million (AIUC) to 25 million (Armilla) to around 9 million (Testudo), with Corgi's structure being different in kind. Three of the four bundle audit and insurance in the same firm. Almost all specialist capacity in the category is US or UK domiciled and writes on London or excess-and-surplus paper, which leaves the European and Asian markets comparatively thin. And the deepest, most durable moat in insurance is not the software. It is accumulated loss data, reinsurer trust, and underwriting judgment, none of which a competitor can scrape from a quote form.

“three of these four grade the homework they then insure.”

The gap this map reveals is independence. That is defensible in a cold-start market with no loss data, and it may not survive the market's maturity. goodfault's position is the deliberate opposite: recognize any credible third-party audit, from AIUC-1 to ISO 42001 to an independent red-team report, and never run the audit ourselves. We think the buyer is better served when the party accepting the certificate has no stake in the grade, and we think that is where the category is heading.

Pick the structure that fits your situation. If you want certification and cover in one motion, the bundlers are built for that. If you want speed from litigation data, Testudo is built for that. If you want a whole startup stack on one carrier's paper, Corgi is built for that. And if you want an insurer whose only job is to carry your risk, with your safety judged by someone who is not also your insurer, that is the one we built.