Units of Berkshire Hathaway, Travelers Group, and Chubb are seeking to exclude or restrict AI-related damage from standard liability policies as generative AI lawsuits grew 978% in the US between 2020 and 2025, while Corgi Insurance, a YC-backed San Francisco startup that raised $108 million in January 2026 and received regulatory approval as a full-stack licensed carrier, has introduced a dedicated AI liability coverage line designed specifically for companies deploying AI systems in place of human judgment.
The coverage gap the major carriers are creating is specific enough to be commercially significant. Berkshire, Travelers, and Chubb are not cancelling all policies for AI-deploying companies. They are introducing AI-related exclusion language into standard commercial general liability, technology errors and omissions, and professional liability policies, effectively removing coverage for claims that arise specifically from AI system outputs rather than from traditional professional negligence or product defect. The practical consequence for a startup whose AI product makes an incorrect recommendation that causes financial harm to a client is that their existing E&O policy may deny coverage on the grounds that the harm arose from AI output, which is now excluded, rather than from a human professional judgment, which is covered. A Gallagher Re and MIT report published this spring documented that standard cyber, tech E&O, product liability, and commercial general liability policies each leave significant coverage gaps for AI-driven incidents, and that courts are currently treating AI as a tool, placing liability on the organisations deploying it rather than on the model providers. That means the company with the startup's API key is the defendant, not Anthropic or OpenAI.
Corgi's AI liability coverage line is the product built specifically into this gap. The policy covers claims arising from AI errors that cause client financial harm, a coverage category that the established carriers' exclusion language is designed to remove from standard policies. The startup program also provides Directors and Officers coverage, Cyber, Technology E&O, Employment Practices Liability, and Media Liability, all quoted instantly online through Corgi's AI-underwritten platform rather than through a broker-mediated process that takes weeks. Co-founder Nico Laqua, a former Palantir employee, built the carrier with the specific observation that startup insurance procurement is one of the most friction-intensive administrative tasks founders face, consuming weeks of their time for a set of products that most brokers do not understand well and that incumbent carriers have not redesigned in decades. Corgi received a California insurance carrier licence in 2025, backed its launch with $108 million from Y Combinator, Kindred Ventures, and other backers, and has since acquired Corgi.com to signal the seriousness of the brand investment. It is not an insurtech broker redistributing someone else's paper. It is a licensed carrier controlling its own underwriting decisions, which means it can write the AI liability coverage that established carriers are unwilling to touch.
The enterprise sales implication is the angle that matters most for founders in AI-heavy workflows. Enterprise procurement teams at financial services firms, healthcare organisations, and large regulated businesses are increasingly requiring that AI vendors carry specific coverage for model errors and AI liability as a condition of contract. The requirement is not yet universal, but it is appearing in more vendor qualification questionnaires than it was twelve months ago, and the movement by major carriers to exclude AI-related claims from standard policies makes it harder for startups to satisfy that requirement with an existing E&O policy. A startup that can present a Corgi AI liability policy during enterprise procurement is not just complying with a contract requirement. It is signalling to the buyer that its AI deployment is considered insurable, that an independent carrier has evaluated the risk and determined that coverage is commercially viable, and that a claims process exists if something goes wrong. That signal, as AI products move from demo risk into balance-sheet risk, is worth something in sales cycles where trust is the primary bottleneck.
The pricing and exclusions in AI liability policies are the market's most honest statement about what insurers believe AI systems can and cannot safely do at this stage of the technology. Insurance markets price based on expected claims frequency and severity, modified by whatever risk management and governance controls the insured party has in place. A carrier willing to write AI liability coverage is implicitly asserting that the category of risk is modellable, that the expected loss distribution has a finite upper tail, and that governance practices like human review, audit logging, and incident response planning reduce the risk to levels that make the policy commercially viable. The exclusions are equally informative: most AI liability policies exclude intentional misuse, exclude coverage for incidents involving classified or defence applications, and typically impose restrictions on autonomous agent workflows where no human is in the approval chain. Those exclusions are the insurance market's implicit evaluation of where AI systems are not yet governable enough to underwrite. Every exclusion is a category of AI deployment that a major insurer has decided it cannot model.
Aon's AI Risk 2026 report, published earlier this year, documented that more than 90% of insurance decision-makers consider AI-driven incidents a material concern, that AI-generated phishing now achieves click-through rates of 54% against 12% for traditional attacks, and that boards face rising Directors and Officers scrutiny over AI governance decisions specifically. That last point is the one most likely to drive enterprise procurement requirements for AI liability coverage beyond what startup founders are currently anticipating. When a company's board faces personal D&O exposure for the AI governance decisions its management makes, the board's interest in ensuring that AI liability coverage exists at the operating level becomes direct rather than theoretical. The insurance market is not yet fully built for this transition, but Corgi's AI liability line, the major carriers' exclusion language, and the Gallagher Re and MIT documentation of coverage gaps are all signals that the transition is now commercial, not speculative. Founders selling AI automation into regulated enterprises should treat AI liability insurance as a sales enablement requirement rather than a compliance formality.
Also read: Greg Brockman Just Confirmed OpenAI Is Exploring an IPO Under Oath and the Implications Run Much Deeper Than the Trial Headline • Open-Weights Image Models Are Narrowing the Quality Gap with Paid Frontier APIs and the Founder Build-Versus-Buy Calculus Is Shifting Faster Than Most Roadmaps Assume • DoorDash Just Embedded AI Into the Entire Merchant Onboarding Stack and That Changes the Math for Startups Serving Local Restaurants
