| Armilla | Testudo | |
|---|---|---|
| Built for | Companies building and deploying AI models and products. | Companies deploying generative AI in their operations and products. It excludes companies selling AI products. |
| Form | Standalone AI liability policy from a Lloyd's coverholder whose sole focus is AI liability. | Standalone, claims made GenAI liability policy from a Lloyd's coverholder. |
| Limits | $25M aggregate per insured, $5M per model, raised January 2026. | Published capacity was $9.25M per insured in March 2026 materials; later trackers cite $10M aggregate. The figures conflict, so confirm at submission. |
| Capacity | Chaucer (lead), AXIS Capital, Convex at Lloyd's. | Apollo, Atrium, and QBE at Lloyd's. |
| Covers | Model error and drift, hallucinations, harmful outputs, data leakage, bias, defence costs, insurable regulatory fines (EU AI Act and Colorado AI Act named). | Output errors, IP and defamation from AI content, unauthorised data disclosure, bodily injury and property damage from AI driven decisions, regulatory violations. |
| Distribution | Brokers only. | Brokers only. |
| Where Helix fits | Accessed via our wholesale partners; the comparison between the two is the advice. | Accessed via our wholesale partners; choosing wrongly between builder and deployer cover is the most common AI placement error we see. |
Our verdict
Build AI for customers: start with Armilla and your Tech E&O wording. Deploy GenAI inside a non AI business: start with Testudo. Both are broker only, so the route to either is a broker, not a checkout page.
Comparison compiled from published materials in our research, checked November 14, 2026. Company claims that could not be corroborated outside a company's own channels are described as claims.
Talk through your stack before you sign the next contract
Send us the term sheet, the customer MSA, or the renewal you are worried about. We will tell you plainly what it requires, what your current programme does about it, and what we would place differently.