Short answer, verified 17 August 2026. No European-domiciled carrier sells an off-the-shelf AI agent liability policy. Four sources of dedicated cover reach a European buyer today: Munich Re aiSure through Mosaic, up to EUR/USD/CAD 15 million; Armilla, a Coverholder at Lloyd's, up to USD 25 million per organisation; Counterpart, on professional liability lines; and bespoke Lloyd's programmes placed through a broker. Your existing errors and omissions, cyber and general liability policies probably do not respond; the three-line coverage framework shows where each one stops. The full carrier-by-carrier picture, with the date every line was verified at the carrier's own domain, is in the carrier matrix.
A formal registry for organisations preparing to insure autonomous AI systems ahead of the Union's enforcement deadlines. Coverage opens in alignment with the Artificial Intelligence Act and the revised Product Liability Directive.
The EU AI Act high-risk obligations may shift to 2 December 2027 if the Digital Omnibus is adopted. The insurance market does not pause. Capacity, exclusions, and product launches continue on their own clock. The PLD transposition deadline of 9 December 2026 is unchanged.
AI Act adopted. Operator provisions enforceable on 02.08.2026. PLD 2024 follows on 09.12.2026.
AIUC-1 precedent only. Reinsurer programmes active. Primary market pending.
ISO 42001, NIST AI RMF, AIUC-1, Agent Certified methodology publicly available.
Most European operators have not yet audited their agent stack against Article 26 obligations.
Figure A. The Agent Insured Market Readiness Index. Composite of regulatory activation, carrier activity, standards maturity, and operator preparation. Updated weekly as the European market approaches the August 2026 deadline. Methodology available on request.
Every specialist carrier writing meaningful AI liability limits in 2026, in one interactive reference. Filter by segment, geography, and product posture. Updated quarterly. Inclusion is editorial only, no carrier pays for placement.
From August 2026, the European Union treats the deployment of autonomous AI systems as a regulated activity. From December 2026, the harms they produce sit inside product liability law. Together these two instruments turn AI agents into an insurable class of risk, and expose every operator that has not secured coverage.
Providers and deployers of general-purpose AI systems, and operators of high-risk AI under Annex III, face conformity, transparency, and post-market monitoring obligations. Administrative fines reach EUR 35 million or 7% of global turnover.
Software, including AI systems and the data they rely on, is formally treated as a product. Claimants gain disclosure rights and a rebuttable presumption of defectiveness where an AI system is shown to have contributed to damage.
Underwriting is anchored on certification evidence and deployment telemetry. Organisations registered with the pre-launch desk are invited into binding quotation in the order of their registration.
An AI agent policy is not a single cover. It is a composite of three risk pillars that together meet the evidentiary expectations of Article 26 of the AI Act, Article 10 of the revised Product Liability Directive, and the four underwriting questions every European AI insurer has started to ask. The seven Agent Certified dimensions feed the pillars. The pillars compose the policy.
Underwriter scrutiny on model reliability, evaluation coverage, failure modes and provenance of data the agent reasons over. The most scrutinised pillar at submission.
Board-level accountability, oversight staffing, role definition, and the autonomy envelope the operator has set for the agent. The pillar regulators focus on.
Who invokes the agent, under what authority, with what approval and rollback controls. The pillar that defines first-notice-of-loss triggers.
No policies are being sold today. This is the indicative framework insurers have begun to price against. An Agent Certified tier discount reduces the loading that autonomy and sector exposure introduce, which is the single largest mechanism operators can use to move the number down before the Q3 2026 coverage window.
An AI system that can act on its own behalf will eventually produce a loss that nobody expected. The work of the coverage market is to decide, in advance, who carries that loss.
Organisations registered with us are notified in advance of underwriting guidance, receive the weekly Agentic Liability Monitor, and are invited into binding quotation before general availability.
Long-form notes on how AI agent liability is being priced, what the emerging insurance standards actually cover, and how European enterprises should prepare for the August and December 2026 deadlines.
The Digital Omnibus deferred Annex III to December 2027. Here is why European AI agent insurance pricing and demand did not move with it, confirmed August 2026.
EU AI Act Annex III makes hiring, promotion, termination, and worker-monitoring AI high-risk. Why EPL insurance was not built for it, and what closes the gap.
Certification helps pricing but is not a precondition for cover. Here is what European carriers actually accept from an uncertified AI agent in 2026.
The EU AI Act's high-risk deadline has passed. Here is what actually changes for AI agent insurance buyers now that Article 26 and Article 99 are live law.
Every article published by Agent Insured on AI agent liability insurance in Europe. 47 pieces on carriers, coverage grants, exclusions, underwriting evidence and claims triggers.
Article 99 fines reach EUR 35 million or 7 percent of turnover. Whether an AI liability policy can pay them is a question of public policy, not just wording.
When your AI agent transacts with another company's AI agent and the deal goes wrong, does any current AI insurance product respond. A market read for 2026.
Thousands of your clients could rely on the same foundation model at once. Whether a single AI failure across many policyholders is covered depends on aggregation wording.
When an AI agent calls an external tool or plugin that fails, coverage depends on policy wording most enterprises have not checked. What to verify before you assume.
AIG, Great American and WR Berkley filed to broaden AI exclusions in E&O, D&O and cyber policies. What it means for European buyers under US-parent programmes.
The Omnibus delay landed and AI underwriting did not move. Why specialist capacity priced through the whole uncertain window, and what the sixteen months to 2 December 2027 are actually for.
Do European enterprises need to disclose AI agents when applying for coverage? A guide to disclosure duty, material fact, and non-disclosure remedies in 2026.
What sublimits and aggregate caps actually restrict in AI agent insurance policies, how they interact with the five trigger categories, and how to read them.
If your high-risk AI deployment lacks a completed Fundamental Rights Impact Assessment under Article 27, will your AI liability insurer still pay a claim? Here is the answer.
Agent Liability EU is the operator desk on those instruments. Read the obligations that put this coverage on the European agenda.
agentliability.eu → The methodologyAgent Certified is the published methodology that feeds the Coverage Architecture above. Review the scoring rubric before submitting a registration.
agentcertified.eu →