Cowbell will sell a standalone policy from January 1, 2027 covering losses when AI agents used by small businesses make mistakes, move money wrongly or cause third-party claims. It arrives as 2,369 generative AI exclusion filings spread across 49 states.
Cyber insurer Cowbell has announced what it calls the first standalone liability policy built for small and mid-sized businesses that let AI agents act on their behalf. Cowbell AI Liability Insurance will go on sale on January 1, 2027, through wholesale brokers on a non-admitted basis, which means it is written outside the standard state-approved forms.
What changed
AI agents are different from chatbots. They carry out tasks with little or no human approval: answering customers, issuing refunds, paying invoices, booking appointments or filing documents. When an agent makes a mistake, such as sending money to the wrong account or giving a customer wrong information, the loss can fall between existing policies.
Cowbell's policy uses a new trigger, an "AI incident," rather than the security breach that usually triggers cyber cover. According to the company, it covers first-party costs such as investigating and containing an incident, restoring operations, and financial losses from transactions an agent executes in error, as well as third-party claims from customers or others harmed by an agent's output. It also covers LLMjacking, where outsiders use stolen credentials or API keys to run up AI usage charges on a company's account. Cover is subject to limits, retentions and exclusions, and Cowbell has not published premiums.
Why it matters now
Standard policies are writing AI out. A review of state filings reported by Insurance Business found 2,369 generative AI exclusion records in force across 49 states by mid-2026, mostly in general liability forms, with no admitted standalone product to fill the gap. That leaves many businesses with less cover than they think as they adopt AI tools.
Cowbell says it will price the risk using three AI-specific factors that grade how an agent is monitored and constrained: for example, whether a human approves high-value actions, what systems and money the agent can reach, and how its activity is logged. Trent Cooksley, Cowbell's co-founder and chief operating officer, said small businesses are giving AI more responsibility for how they operate. Simon Hughes, chief commercial officer, said the product gives brokers a concrete solution for clients' agent risks.
What it means for small businesses
Any business using AI agents for customer service, payments or operations should check three things before renewal. First, read the general liability and cyber policies for AI exclusions, which may have been added at the last renewal. Second, ask whether a loss caused by an agent's mistake, with no hacker involved, would be covered anywhere. Third, look at controls: limits on what an agent can pay or approve, human sign-off for large transactions, and logs of what the agent did. Those controls reduce the chance of a loss and will likely affect the price of any AI cover.
Because the policy is non-admitted, it is not covered by state guaranty funds if the insurer fails, and its wording is not pre-approved by regulators. Businesses should ask a broker to compare the wording with any AI endorsements available from their existing insurers.
The wider market
Insurers have been cautious about AI because losses are new and hard to model. A handful of specialist products have appeared for large firms, often covering errors by AI models in specific uses. Cowbell's move into small businesses is notable because that is where AI agents are spreading fastest and where in-house risk management is thinnest. Expect more insurers to follow if early claims experience is manageable, and expect underwriters to ask more detailed questions about AI use at every renewal.
Key Points
- 1Cowbell AI Liability Insurance goes on sale January 1, 2027 through wholesale brokers on a non-admitted basis.
- 2It is triggered by an 'AI incident', not a security breach, and covers first-party losses and third-party claims caused by AI agents.
- 3Cover includes errors in transactions an agent executes and LLMjacking charges from stolen API keys.
- 42,369 generative AI exclusion filings were in force across 49 states by mid-2026.
- 5Pricing uses three factors that grade how agents are monitored and constrained.
Why This Matters
Small businesses are adopting AI agents faster than their insurance is adapting, and many standard policies now exclude AI. Knowing where the gaps are helps owners avoid an uninsured loss.
Understand the numbers
- GuidePercentage Deductibles Explained: Wind, Hail and Hurricane Cover
- GuideWhat Is a Combined Ratio? How to Read an Insurer's Earnings in Plain English
- GuideHow Much Home Insurance Do You Need? Replacement Cost vs Market Value
- CalculatorInsurance Deductible Break-Even Calculator ยท Compare a low and a high deductible using your own claim odds.
Original Source
Cowbell announcement via PR Newswire; Insurance Business, SiliconANGLE and Insurtech Insights (November 10, 2026) โRelated Stories
Who Should Refinance at 7.11%? Break-Even by Loan Vintage
November 18, 2026
Binance's $6.3bn Hack Shows How Little Crypto Theft Is Insured
November 18, 2026
Australian Home Cover Rises 6% as Rates Hit 4.85%: The Squeeze
November 18, 2026
Microsoft 365 Outage: Why Most of the Loss Will Stay Uninsured
November 17, 2026
Daily Intelligence
The PolicyRix Daily Brief
Get the day's key insurance and finance stories in one email, each linked to its original source. No spam. Unsubscribe anytime.
Informational newsletter only. Not financial advice. Disclaimer