San Francisco-based insurance startup Corgi has launched AI Insurance Coverage, a modular product designed to sit on top of a company’s existing Tech E&O policy. The coverage is built for six AI-related risk scenarios, including algorithmic bias, harmful or incorrect generated content, and disputes tied to training data. The move targets an area many traditional insurers have largely avoided, often excluding AI-caused losses from standard policy terms.
Coverage targets six distinct AI failure scenarios
Corgi said the product currently covers algorithmic bias, harmful or incorrect generative outputs, misuse of training data, adversarial attacks, synthetic media, and failures involving autonomous systems. The examples are practical and business-facing: an AI hiring tool that systematically screens out a certain group, a customer service bot that gives incorrect legal or medical guidance, or an AI agent that takes harmful actions without human oversight.
In Corgi’s framework, adversarial attacks refer to inputs crafted to push a model into producing wrong outputs, while synthetic media includes deepfake-style fabricated content. The company is packaging these risks as modules that businesses can attach to existing coverage instead of replacing their current insurance structure.
$108 million raised and regulatory approval secured
As of January 2026, Corgi said it had raised a total of $108 million. The company also said it had obtained regulatory approval, making it the first full-stack insurer built for startups, with underwriting, policy administration, and claims handling managed in-house rather than backed by traditional insurers.
Since receiving full underwriting authority in July 2025, Corgi said its annual recurring revenue has surpassed $40 million. Co-founder and CEO Nico Laqua said in a statement that companies are moving quickly on AI adoption, while insurance has not kept pace.
AI is changing how liability gets defined
The product arrives as companies place AI systems directly into production workflows for loan reviews, resume screening, and customer support. That shifts the liability debate. A software bug can often be traced to specific code, but a flawed model decision may stem from training data, system behavior, or parameters that are far harder to isolate.
That difference matters for insurers and businesses alike. As AI systems take actions with financial, legal, or reputational consequences, the boundary of responsibility becomes harder to map. Corgi’s new coverage is aimed at that gap, where demand appears to be forming before much of the insurance market has fully responded.