AI liability insurance is its own category now. Whether a standard errors-and-omissions or cyber policy responds to AI-caused losses depends on its wording; Corgi's AI page, for example, says generic tech E&O policies written before generative AI may be silent or ambiguous about model outputs. A new class of insurers writes AI-specific cover for losses such as hallucinated outputs, autonomous agent mistakes, IP infringement, data exposure, and regulatory exposure. Below are the strongest options for companies building or deploying AI, and who each one actually fits.
This is an independent guide from QuoteSweep, which maps the modern commercial insurance landscape. QuoteSweep does not sell any of these companies' products.
TL;DR: Corgi is an AI-native insurance group that bundles Tech & AI liability into startup insurance, written on its own unrated risk retention group paper. Testudo writes a standalone, Lloyd's-backed generative AI policy ($2M sub-limit per insuring agreement, $10M aggregate) through brokers. Armilla pairs affirmative AI coverage with independent model verification and reports limits up to $25M. AIUC combines the AIUC-1 certification standard, audits, and insurance. Klaimee insures autonomous AI agents specifically, covering both first- and third-party harm.
Quick picks
- Best for a full startup insurance stack that includes AI risk: Corgi
- Best standalone generative AI policy (Lloyd's-backed): Testudo
- Best AI liability paired with independent verification: Armilla
- Best certification-plus-insurance standard: AIUC
- Best for autonomous AI agents specifically: Klaimee
The best options, compared
Corgi — best for a full startup stack that includes AI
Corgi is an AI-native insurance group for startups and technology companies. You buy from its licensed producer and the risk is carried primarily by Technology Risk Retention Group, Inc. (or by Corgi's admitted carrier or affiliated or partner carriers), a risk retention group that is not AM Best rated and, per Corgi's own disclosure, carries no state guaranty-fund backstop. It packages modular coverage by funding stage from pre-seed to growth. Its lines include Commercial General Liability, cyber, D&O, EPLI, and fiduciary, plus a Tech & AI Liability (tech E&O) line. AI-related claims are excluded from the standard tech E&O policy and added by endorsement, so AI cover sits inside one stack rather than a separate purchase. Founders can self-serve a quote in minutes and bind the same day. Corgi's press page reports a $160M Series B (May 2026) and a $106M Series B1 at a $2.6B valuation (May 2026).
Best for: funded startups and AI companies that want their whole insurance program — including AI risk — from one counterparty, and for whom an unrated risk retention group is acceptable paper.
Testudo — best standalone generative AI policy
Testudo is a Lloyd's of London coverholder writing a standalone generative AI liability policy — a dedicated claims-made policy, not an endorsement bolted onto an existing form. It covers AI output errors, IP infringement, unauthorized data disclosure, bodily injury, property damage, and regulatory violations. It is written on Lloyd's syndicate paper on a surplus lines basis (Testudo says the A+ (Superior) AM Best rating belongs to Lloyd's, not to Testudo), with a $2M sub-limit per insuring agreement and a $10M aggregate, and it is not available in all states. It covers generative AI systems a company deploys, not ones it builds or licenses. Testudo sells only through brokers and says a complete submission can expect a non-binding indication within 48 hours; it assesses exposure from public sources without access to your systems. For the head-to-head with Corgi, see Corgi vs Testudo.
Best for: US companies deploying generative AI that work with a broker and want dedicated, Lloyd's-backed coverage.
Armilla — best AI liability paired with verification
Armilla offers purpose-built affirmative AI liability insurance for generative AI and AI agents, paired with independent AI verification and risk assessment. The verification side is the differentiator: before or alongside coverage, Armilla assesses the model itself — verification, bias evaluation, stress testing, and red teaming — plus ongoing monitoring. As a Lloyd's coverholder it writes affirmative coverage for model errors and hallucinations, output liability, AI agent mistakes, data leakage, property damage, and regulatory violations, with capacity backed by partners including Chaucer, Axis Capital, Convex, Swiss Re, and Greenlight Re, and reported limits up to $25M per organization. It is distributed through brokers.
Best for: companies that want AI liability coverage paired with an independent view of how the model actually behaves — and the highest reported limits of this group.
AIUC — best certification-plus-insurance standard
AIUC, the Artificial Intelligence Underwriting Company, combines three things: the AIUC-1 certification standard for AI agents (six pillars — data and privacy, security, safety, reliability, accountability, and societal risks), independent third-party audits performed by accredited firms (Schellman is noted as the first accredited auditor), and insurance against losses from AI agent failures, priced by how safe a system is shown to be. AIUC-1 draws on established frameworks including MITRE ATLAS, and the certification doubles as a trust credential a vendor can show a cautious enterprise buyer. Per reporting, AIUC emerged from stealth in 2025 with a $15M seed backed by Nat Friedman.
Best for: enterprises and AI vendors that want a certified-and-insured approach, where a formal, audited standard becomes an enterprise-trust credential.
Klaimee — best for autonomous AI agents specifically
Klaimee is a Y Combinator-backed startup providing liability insurance and certification for autonomous AI agents, covering AI agent losses. It insures both third-party harm (an agent harming your customers) and first-party harm (an agent damaging your own systems, data, or records) — for scenarios like hallucinated outputs, unauthorized actions, data exposure, prompt-injection attacks, and wrongful communications. Coverage is paired with an audit that scores an agent's risk across eight dimensions and issues a certification, plus ongoing access to an adversarial testing environment. It focuses on B2B companies with bounded use cases (not autonomous vehicles, robotics, or physical-safety systems).
Best for: AI vendors and enterprises insuring autonomous agents, especially those that want first-party damage covered alongside third-party harm.
At a glance
| Company | Best for | Model | Notable |
|---|---|---|---|
| Corgi | Full startup stack incl. AI | Insurance group: producer + affiliated RRG (unrated) | Tech & AI liability line; quote in minutes and same-day bind on core lines; publishes planning price estimates |
| Testudo | Standalone generative AI | Lloyd's coverholder / MGA (A+ rating belongs to Lloyd's) | $2M sub-limit per insuring agreement, $10M aggregate; surplus lines, not in all states; broker-only, indication within 48 hours |
| Armilla | AI liability + verification | Lloyd's coverholder | Affirmative coverage; up to $25M reported; Swiss Re, Chaucer among partners |
| AIUC | Certified + insured standard | Standard + audits + insurance | AIUC-1 six pillars; accredited audits (Schellman); $15M seed, Nat Friedman |
| Klaimee | Autonomous AI agents | Insurance + certification | First- and third-party harm; 8-dimension agent audit; YC-backed |
How to choose
- Start with what you're insuring. Building or deploying autonomous agents specifically → Klaimee. Generative AI you deploy → Testudo. Your whole startup's risk, with AI included → Corgi.
- Decide whether you want verification or certification bundled. Independent model verification and red-teaming → Armilla. A formal, audited standard and enterprise-trust credential → AIUC. An agent-level audit plus ongoing adversarial testing → Klaimee.
- Weigh who's behind the policy and the limits. Lloyd's-backed capacity (the A+ rating belongs to Lloyd's) → Testudo and Armilla. Corgi owns underwriting and claims but writes primarily on an unrated risk retention group, so check whether your contracts require rated, admitted paper. Highest reported limits (up to $25M) → Armilla; $2M sub-limit per insuring agreement and $10M aggregate → Testudo.
- Consider distribution and speed. Self-serve with same-day binding → Corgi. Broker-placed, with an indication within 48 hours of a complete submission → Testudo. Broker-distributed → Armilla. A short declarative application plus an agent audit → Klaimee.
- Compare quotes on the same terms. None of these companies publish flat pricing (Corgi publishes planning estimates), so pricing has to be quoted for your specific system and coverage.
Compare Corgi and Testudo directly in Corgi vs Testudo, and compare all of them side by side on the ai-agents hub.
Frequently Asked Questions
What is AI liability insurance?
It is coverage for losses caused by AI systems — hallucinated or negligent outputs, autonomous agent mistakes, IP infringement, unauthorized data disclosure, property damage or bodily injury from reliance on AI, and regulatory exposure. Whether a standard general liability, E&O or cyber policy responds to AI-caused losses depends on its wording, and these specialist insurers write cover aimed at AI-specific risk. For the broader category, see AI agent insurance explained.
Which is best if I deploy generative AI?
Testudo writes a standalone, Lloyd's-backed generative AI liability policy ($2M sub-limit per insuring agreement, $10M aggregate) for companies that deploy generative AI. It sells only through brokers, is not available in all states, and gives a non-binding indication within 48 hours of a complete submission. Armilla is the alternative if you also want independent model verification bundled with the coverage, and reports higher limits (up to $25M).
Which is best for autonomous AI agents specifically?
Klaimee is purpose-built for autonomous agents, covering both third-party and first-party harm and pairing coverage with an eight-dimension agent audit. AIUC suits teams that want a formal certified standard (AIUC-1) plus insurance, and Armilla covers AI agents alongside generative AI with independent verification.
How much does AI liability insurance cost?
None of these companies publish flat pricing, though Corgi publishes planning estimates. Cost depends on your system, coverage, and — for the assessment-led insurers — how safe your AI is shown to be. AIUC, for example, ties pricing to how safe a system tests, so quotes are specific to each deployment.
The bottom line
The best AI liability insurance depends on what you're insuring and how much assurance you want bundled: Corgi for a full startup stack that includes AI, Testudo for a standalone Lloyd's-backed generative AI policy, Armilla for coverage paired with independent verification and the highest reported limits, AIUC for a certified-and-insured standard, and Klaimee for autonomous AI agents specifically. Compare them side by side on the ai-agents hub, or read AI agent insurance explained for the full category.

