The premise sounds almost too current to be real: What happens when your autonomous AI agent makes a promise your company can't keep, deletes critical data, or says something actionable to a customer?
Until recently, the answer was murky at best. Traditional liability policies weren't designed for software that makes its own decisions. Now a San Francisco startup thinks it has found the gap—and the market.
Klaimee, fresh from Y Combinator's Spring 2026 cohort, announced $5.5 million in seed funding on July 22 to launch what it's calling dedicated liability insurance for autonomous AI agents. The timing isn't coincidental. As major carriers have begun adding AI exclusions to standard commercial policies, companies deploying these systems are discovering they may not be covered for what happens next.
Following the Exclusions
FundersClub led the round, joined by a cluster of venture firms and angel investors including ex/ante, Pioneer Fund, Multimodal Ventures, Kima Ventures, Rebel Fund, Robinhood Ventures, and Y Combinator itself. Valuation remains undisclosed, as does the specific A-rated carrier providing Klaimee's insurance capacity—a detail the company isn't ready to share.
But the real story may be less about Klaimee's funding and more about what's happening in the broader insurance market.
Starting January 1, 2026, the Insurance Services Office introduced three new exclusion endorsements—CG 40 47, CG 40 48, and CG 35 08—that carve generative AI risks out of standard commercial general liability policies. The forms began appearing on renewal notices across the country. Verisk has hinted that additional language targeting agentic AI may be coming.
The result? A coverage void that traditional policies can't quite fill. Cyber insurance typically addresses external attacks by human threat actors. Errors and omissions policies cover bugs in software written by developers. An AI agent that autonomously decides to offer a refund you never authorized, or makes a false statement in a debt collection call, falls somewhere between those lines.
Perhaps more than insurers expected.
Testing Before Insuring
Klaimee's approach pairs certification with coverage. Companies submit their AI agents to what the startup describes as more than 100 behavioral probes spanning eight risk categories: scope violation, data exfiltration, unauthorized actions, output integrity, adversarial manipulation, behavioral stability, model drift, and operational control failures.
The testing process, which Klaimee says takes about ten minutes to initiate, returns a letter-grade risk score within days along with a remediation roadmap. Pass the evaluation and your agent earns a "Klaimee-Verified" badge—and becomes eligible for AI-specific liability coverage, with premiums keyed to the certification score.
Coverage splits into two buckets. Third-party liability applies when an agent harms someone outside the organization. First-party loss kicks in when it damages the deploying company itself. Examples run from AI voice agents making legally dubious collection threats, to customer support bots promising services that don't exist, to internal automation systems wiping production databases.
Claims get processed through parametric triggers tied to system logs and operational data, a design intended to speed payouts compared to traditional investigative processes. For now, Klaimee is sticking to B2B software use cases—explicitly steering clear of autonomous vehicles, robotics, or anything that could cause physical harm in the real world.
A Narrow Market, for Now
Klaimee isn't alone in spotting the opportunity. Mayflower Specialty, an AI-focused managing general agent, began writing standalone AI liability policies in May with initial limits reaching $5 million. Some established carriers, including Canopius US and Allianz Commercial, maintain that existing policy structures can absorb AI exposure, sometimes with sub-limits tacked on.
Whether the market is large enough to sustain multiple entrants remains an open question. AI agents are moving into production faster than many enterprises anticipated, but the operational footprint is still relatively contained—customer service bots, internal workflow automation, limited decision-making roles.
That could change quickly. Or not.
The Operator
CEO Ines Boutemadja comes to Klaimee with a specific résumé. She previously ran operations at SafetyWing, where she built insurance products from the ground up, structured an MGA with Tokio Marine, secured a Puerto Rico carrier license, and helped scale revenue from $5 million to $60 million. She holds degrees from HEC Paris and the National University of Singapore and notes that she's the first Algerian woman founder to go through Y Combinator—a distinction that matters in an industry not known for demographic diversity.
CTO Julien Catonnet's background spans engineering, corporate strategy, and enterprise compliance. The company lists between two and ten employees, though the exact count may not reflect recent hires.
What Comes Next
The seed capital will fund the rollout of Klaimee's insurance-backed warranty program. The startup wrapped a public "adversarial testing playground" on July 18, offering free evaluations to early participants—a marketing move as much as a technical one.
The product seems calibrated for a specific buyer: enterprise procurement teams. Certification PDFs and verified badges aren't primarily technical artifacts; they're procurement artifacts. Legal departments and risk managers need something to attach to vendor reviews, and as AI agents shift from experimental tools to production systems handling real customer interactions, demonstrable coverage becomes less optional.
Traditional policies, with their new exclusions, may no longer provide that comfort. Whether Klaimee's model fills the void—or simply highlights how strange the insurance landscape has become for autonomous software—will depend on how quickly enterprises adopt agentic systems.
And what those systems do when no one's watching.
