Cyber insurance provider Cowbell Cyber Inc. today unveiled Cowbell AI Liability Insurance, a standalone policy for U.S. small and midsized businesses that covers losses caused by the artificial intelligence agents they deploy.

The new policy is meant to deal with a problem that arises when businesses let agents move money or change customer records with nobody signing off first. Coverage hinges on a trigger the company calls an “AI Incident.” The definition takes in model drift and attacks that manipulate an agent, not just software acting without approval.

Cowbell describes an accounts payable agent misreading approval statuses and paying a supplier twice with no human check. Part of that money disappears when the supplier goes into liquidation.

Unrecovered funds like those can be covered, and the policy also pays to investigate what happened and take the agent offline. Inaccurate or unsafe output can be corrected or withdrawn at the insurer’s expense, with recipients notified.

Usage charges an agent racks up without authorization are covered too, such as when a loop goes unchecked. Cowbell also covers LLMjacking, where an outsider uses stolen credentials or application programming interface keys to run AI workloads and leaves the business holding the bill.

Claims brought by customers fall on the third-party side. Cowbell’s example there is a bookkeeping firm whose agent files client tax returns using outdated rates, and the clients sue once the penalties and interest arrive.

In a case like that, the policy can pay defense costs and any damages or settlements, and the same goes for suits alleging algorithmic discrimination or bodily injury caused by reliance on AI output. Regulatory proceedings are also covered, and where the law allows it so are certain fines and penalties.

Cowbell Chief Commercial Officer Simon Hughes said AI liability is “no longer only about people using AI,” pointing to agents that act with no human in the decision-making chain. When one goes rogue, he said, a business can take its own losses and face customer claims at the same time. Hughes said the policy was written with that double exposure in mind.

Underwriting draws partly on three AI-specific Cowbell Factors covering observability, autonomy and governance. They show underwriters how closely a business monitors its AI tools and how much authority those tools have to act on their own.

Cowbell’s Risk Advisor agent, launched in October, uses the factors to explain risk signals to policyholders. Businesses can see their own scores by connecting Gemini, Claude, ChatGPT and other tools through Cowbell Connectors.

Cowbell already sells some AI coverage. Prime One, which it launched in April for U.S. companies with $250 million to $1 billion in annual revenue, includes affirmative coverage for AI-related incidents inside a cyber policy. The new product is sold on its own and aimed at smaller businesses.

Cowbell AI Liability Insurance goes on sale Jan. 1, 2027, and will be available only through wholesale brokers on a non-admitted basis. Cowbell is a venture capital-backed company that has raised more than $200 million, including a $60 million Series C round from Zurich Insurance Group Ltd. in July 2024.

Photo: Cowbell