You deployed an agent that books, buys, promises or decides on your behalf. If it gets something wrong, your existing policies almost certainly do not respond. This site is the plain guide to what that means and what to do about it.
If you run an SME and you turned on an AI agent this year, you are probably sitting inside all three of these gaps without realising it. Here is the short version. The rest of the site walks each one in detail.
Not theoretical harm. An Air Canada chatbot invented a bereavement discount and a tribunal ordered the airline to pay it. A New York firm filed fictional case citations generated by an AI tool and was sanctioned. Agents that transact, communicate or advise create exposure the moment they go live.
Most Errors and Omissions wordings, cyber policies, and general liability contracts were written before autonomous systems existed. Insurers are now adding explicit AI exclusions at renewal. Even without an exclusion, proving the event was covered gets difficult fast when the decision was made by a model.
New insurance products are coming online in 2026. AIUC-1 just became the first certification standard to underwrite an AI agent deployment. The EU AI Act and the revised Product Liability Directive are pushing the market to respond. Coverage exists. You have to prepare for it.
Before you talk to a broker, a lawyer, or a certification body, you need to know the answers to three questions. We wrote them as a short diagnostic. It takes about ten minutes to read and will tell you exactly what to do next.
Read the three questionsDocumented scope, guardrails, and human checkpoints.
E&O, cyber, general liability, D&O. All four.
Who pauses the agent, who calls legal, who notifies users.
If you think this is too early to worry about, the regulators and insurers disagree. Two developments in the last eighteen months are worth watching closely.
“The first AI agent deployment to be underwritten against a formal standard was certified against AIUC-1 in 2025.”
AIUC-1 is a certification and assurance standard for AI agents developed by the AI Underwriting Company with participation from Munich Re, one of the largest reinsurers in the world. The headline case involved an ElevenLabs voice agent deployment. It is the first time an insurer has accepted a structured AI evaluation as the basis for underwriting an autonomous system. Read the full context.
On the regulatory side, the EU AI Act entered into force in 2024 with the first operator obligations starting August 2, 2026, and the high risk obligations following in 2027. The revised EU Product Liability Directive explicitly brings software and AI systems inside strict liability rules. If your agent is operating in Europe, the legal standard of care has already changed.
Written for operators and founders in plain English. Each one sits on a specific question we keep hearing from people running AI in production.
Your E and O, cyber, and general liability policies were written before autonomous agents existed. Here is what they actually say when things go wrong.
Case studyA tribunal decision worth less than a thousand dollars rewrote the liability picture for every business running a customer-facing agent. Here is the full story.
Operator checklistA short diagnostic for founders and operations leads who want to move fast without shipping an exposure they cannot defend.