There is no magic here. Insurers underwriting AI agent risk in 2026 need structured evidence: a documented deployment, an independent assessment, and an operator who can demonstrate a working response plan. This page tells you how to produce all three, and how to get on the list when the policies start writing.
You cannot insure what you cannot describe. The first step is a structured assessment of every agent you run, against one of the emerging certification frameworks. The most developed right now is AIUC-1, which Munich Re has accepted as an underwriting input. Future Proof Certified provides a seven-dimension evaluation covering governance, context, distribution, trust, product, AI integration, and autonomy.
The assessment output is an evidence package: what the agent does, how it is governed, what could go wrong, what controls are in place, and where the gaps are. Insurers read this before they quote. Regulators ask for pieces of it under the EU AI Act. It is the one artefact that does double duty on every track.
Go to agentcertified.eu for the methodology and to begin the assessment.
AI agent liability coverage is not a shelf product yet. It is being built right now by a small number of specialty carriers and managing general agents, with reinsurance support from Munich Re and others. The first binding policies in Europe are expected in the third quarter of 2026. The carriers writing them are being selective, and they are starting with operators who already have certification in hand.
The waitlist on agentinsured.eu is the entry point. When you register, you will be asked for your certification status, the type of agent you run, the sector you operate in, and your headline revenue figures. You do not need to have completed certification to join, but you do need to have started. The waitlist position informs quote order, not acceptance.
Go to agentinsured.eu to join the waitlist.
When your certification report is complete and the waitlist opens to quote, the carrier will ask for three things: the certification evidence, a live demonstration of your incident response plan, and a short attestation from the board or senior management. The first is already in hand. The second is a tabletop exercise you can run in a morning. The third is a one-page document your legal counsel can draft from template.
None of this is exotic. It is the standard evidence package every specialty underwriter has always asked for, adapted to autonomous systems. The reason it feels unfamiliar is that most operators have never had to produce it for software before. By this step you will have.
This step happens between you, your broker, and the carrier. We do not sit in the room.
The pathway takes eight to twelve weeks end to end for an SME running a small number of agents. If you start in April you are comfortably positioned for the first wave. If you start in August you are still fine for the second wave in early 2027. If you start after your first incident, you are not fine.
Go back to the three questionsCertification work begins. Evidence packages drafted.
European specialty carriers expected to bind first operators.
Existing E&O and cyber policies start adding explicit AI clauses.