When procurement teams at large enterprises began requesting proof of AI agent liability coverage earlier this year, they encountered something curious: silence. Most vendors couldn't produce it. Their existing errors-and-omissions policies either carved out AI failures entirely or offered the kind of vague language that makes corporate counsel nervous.
Klaimee, a two-person insurtech out of San Francisco that recently emerged from Y Combinator, thinks it has spotted an opening. The company launched in April with what it bills as insurance purpose-built for AI agents—a product pairing technical red-team assessments with underwritten liability coverage. "You deploy agents, we cover you," goes the tagline. It's a direct pitch to AI vendors fielding awkward conversations with enterprise buyers who want certificates explicitly naming agent-related risks.
Whether that's a breakthrough or just clever positioning in a suddenly crowded field is the question Klaimee now has to answer.
What Traditional Policies Miss
The product targets two failure modes that slip through conventional coverage. First, third-party liability: think of a voice bot used for debt collection that misstates an account balance, or a customer-support agent that promises a refund the company never authorized. Second, first-party operational damage—corrupted records, deleted production databases, the sort of mess that happens when an internal agent makes a bad call.
Klaimee's policy covers hallucinated outputs that customers act on, unauthorized agent actions, data exposure, prompt injection attacks, and the investigatory costs that pile up afterward. What it doesn't cover: anything that might physically hurt someone. Autonomous vehicles, robotics, industrial systems—all off the table. The focus is squarely on software agents navigating customer service queues, enterprise workflows, and operational back-ends.
Traditional cyber insurance, Klaimee argues, was built around external threats—hackers, ransomware gangs. Tech errors-and-omissions policies assume code written by humans, not agents that drift or adapt over time. In a blog post published in May, the company put it bluntly: when an enterprise customer asks for "AI liability coverage," most vendors hand over a standard certificate that doesn't actually name the new risks everyone's worried about.
Certification as Underwriting
To issue a policy, applicants go through what the company describes as a 10-minute intake, followed by a multi-day evaluation. That includes 30 governance questions and more than 100 adversarial probes—tests for prompt injection, jailbreaks, decision drift, data leakage, biased outputs. The end result is a letter grade across eight risk dimensions: scope violation, data exfiltration, unauthorized action, output integrity, adversarial manipulation, behavioral stability, model drift, operational control.
Only agents that pass certification qualify for the insurance itself. They get a badge, a procurement-ready PDF, remediation recommendations, and—crucially—a financial guarantee. Klaimee claims it can deliver a policy and certificate of insurance in under 24 hours for early adopters, though specifics around limits, deductibles, and pricing remain undisclosed.
So do the carriers. CEO Ines Boutemadja mentioned in a May LinkedIn post that the company was in early-stage capacity conversations with U.S. sponsoring carriers and Lloyd's syndicates. That opacity isn't unusual for an early-stage insurtech, but it does leave open questions about how much underwriting capacity Klaimee has secured and whether those conversations have closed.
A Contested "First"
Klaimee markets itself as the first insurer for AI agents, though ElevenLabs launched AI-specific coverage prior—a bespoke vendor arrangement announced in mid-February for its voice agents, not a third-party product. HSB, a Munich Re subsidiary, rolled out AI liability insurance for small businesses in March, covering losses where general liability policies now exclude AI. Mount, another company from the same Y Combinator cohort as Klaimee, describes itself as an "AI Agent Insurance Carrier" and has been profiled as offering the first purpose-built policy for autonomous agents.
The competing narratives say something about how quickly this market is taking shape. In January, the Insurance Services Office introduced generative AI exclusions on commercial general liability forms. Legal analysts warned recently of the "end of silent AI"—the industry term for risks that policies used to cover by default but no longer do. Insurers including AIG, WR Berkley, and Great American have reportedly sought regulatory approval to carve out AI-related claims, citing fears of correlated, systemic risk.
That fragmentation is creating space—perhaps more than the founders expected—for specialized players willing to underwrite what traditional carriers increasingly won't.
Regulatory Deadlines and Enterprise Demand
Klaimee's timing appears deliberate. The EU AI Act enforcement begins in August, adding regulatory urgency to what was already becoming a procurement headache. The company opened what it calls a "public window" for free adversarial testing through mid-June, a signal it's racing to onboard early customers before compliance pressure intensifies.
Boutemadja, who previously worked on insurance infrastructure at SafetyWing (another Y Combinator alum), framed the product in a May podcast as a direct response to enterprise buyers demanding a financial backstop when agents inevitably fail. The company's early blog content—five posts published between mid-April and late May—reads like a field guide for nervous procurement officers: What language needs to be on the certificate? How do you demonstrate third-party validation?
It's practical stuff, the kind of content you publish when you're selling to people who need to check a compliance box.
An Open Question
Whether Klaimee's model—certification tied to insurance—becomes the industry standard or simply one approach among many is still unclear. But the sheer number of specialized insurers, managing general agents, and insurtechs now entering the space suggests the coverage gap is real. Traditional carriers are retreating. New entrants are wagering they can build underwriting models around red-teaming, real-time evaluation, and agent-specific risk scoring.
For now, Klaimee is betting that enterprises will pay for the certainty a policy provides, even if the details of what that policy covers—and who's backing it—remain somewhat opaque. In a market this nascent, being early might matter more than being perfectly transparent.
