AI agents are getting jobs at banks and insurance companies. The question nobody has fully answered is who pays when one of them makes a bad call.
Harbor, a San Francisco-based insurtech startup, thinks it has an answer. The company launched publicly on October 13, 2026, with a $10 million seed round to build a platform that manages AI agent risk and backs it with dedicated insurance.
What Harbor is building
Base10 Partners, Abstract, and 8VC led the round. Restive and Innovation Endeavors also participated.
The pitch has two halves. The first is a risk management platform, and the second is insurance coverage built specifically for AI agent failures.
It starts by building an inventory of the AI agents a company is running, which is harder than it sounds once agents start multiplying across departments.
From there, Harbor assesses each agent’s operational, security, and regulatory risk in real time. If an agent is about to make a decision that breaks compliance rules, the platform is designed to step in with automated interventions before the damage lands.
The monitoring doesn’t stop after setup. Harbor says its platform reassesses risk whenever the underlying models, tools, or data change.
The insurance half is meant to define liability clearly. When an agent fails, the policy is supposed to make it obvious who is on the hook and what is covered.
Why the timing works
Harbor is aiming at a gap that traditional insurers have been widening on purpose. Policies covering areas like professional liability and cyber threats have started excluding AI-specific risks.
Beginning around January 2026, many US insurers started adding generative AI exclusions to general liability policies. More than 2,300 AI-exclusion filings have been recorded across 49 states.
Harbor is targeting highly regulated sectors, specifically banks and insurers across North America, Europe, and Latin America. These are the institutions where a rogue automated decision carries regulatory consequences on top of financial ones.
The company was founded by Jesús Santos, Daniel Elechiguerra, Cristian Bartolomé, and Luis Botín, who met at Stanford University. The founders argue that quantifying AI agent risk is a prerequisite for getting enterprises to adopt the technology more widely.
What this means for AI adoption in finance
The most interesting thing about Harbor is the customer list it is chasing. Insurers are among its target clients, which means the company wants to help insure the very industry that has been backing away from AI risk.
For banks, the appeal is more defensive. Financial institutions answer to regulators who tend to be unimpressed by explanations like “the algorithm did it.” A system that inventories agents, flags non-compliant decisions, and keeps a running risk assessment gives compliance teams something concrete to show examiners.
There are real challenges, though. Pricing a risk with a short track record is notoriously difficult, and AI agents are changing faster than most actuarial tables can be updated.
Competition is another factor to watch. With more than 2,300 exclusion filings already on record, the coverage gap Harbor is targeting is visible to everyone in the insurance business, including incumbents with far deeper balance sheets.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.