Insurance for AI agents: what covers an agent's mistakes

What is AI agent insurance?

AI agent insurance is liability cover that responds when an AI agent, software that takes actions rather than only answering questions, harms a customer or another third party. In the policies we track, agent risk usually sits inside broader AI liability cover rather than in a separate product.

The examples insurers give are concrete. Armilla lists AI agent failures such as wrong decisions, improper tool use and escalation errors¹. AIUC-backed cover is described around harm such as exposing customer personal data or giving fabricated or incorrect information, for example misquoted advice^{2, 3}.

Agents raise a hard question for older policies. When an agent acts on its own, is that a professional service under E&O, a security failure under cyber, or neither? Wording that names agents answers it; silent wording leaves it to a claims dispute.

Which insurance policies cover AI agent errors?

6 listings in our register name AI agent risk. Ollive's is announced but not yet on sale; the others are confirmed on the issuer's own pages.

• AArmillaArmilla Insured$25Mup to, per organizationSurplus lines brokers, or Armilla underwriting after an AI risk assessment
• CChaucer and ArmillaVanguard AI$25M+AI aggregate, alongside $10M cyberSurplus lines brokers
• AAIUCAIUC-1-backed AI agent insuranceLimit not publishedThrough AIUC-1 certification
• CCorgiModular AI coverage on tech E&OLimit not publishedOnline, direct
• MMayflower Specialty and Hadron InsuranceAI liability programLimit not publishedDirect, through a broker, or embedded through partners
• OOlliveAI liability for AI vendors, with runtime monitoring (announced)Limit not publishedBook a demo with Ollive

Armilla's standalone policy responds to third-party losses from AI behavior, including agent mistakes, and is designed to respond ahead of conventional cyber or E&O wording¹. Vanguard AI, from Chaucer and Armilla, is aimed at organizations deploying AI at scale, including customer-facing agents. It pairs Chaucer's cyber and tech E&O with Armilla's AI policy⁴.

Corgi sells AI modules that attach to its tech E&O policy, and the scenarios on its page include autonomous system failures⁵. The Mayflower Specialty and Hadron Insurance program gives companies deploying AI four modules: AI D&O, AI EPL, AI E&O and an AI difference-in-conditions excess layer over an existing program⁶.

Ollive pairs runtime monitoring of AI agents with planned liability cover for AI vendors. Its own page says the insurance program is being designed, with risk to be carried by a licensed insurer it does not name, and it gives no limit⁷. Trade press had said it targeted a summer 2026 launch⁸.

Who needs AI agent insurance, the vendor or the business using the agent?

Often both, because one agent mistake can produce two claims. The business that deployed the agent faces its own customer, and it may then turn to the vendor that built the agent under their contract.

The listings split along that line. Ollive's planned cover is for AI companies running customer-facing agents in production⁹. Vanguard AI is aimed at organizations deploying AI at scale⁴. AIUC-1-backed cover is aimed at agent vendors and their enterprise customers: AIUC says the policy behind ElevenLabs can underwrite the actions of agents that ElevenLabs' customers deploy¹⁰.

Armilla's policy is written for companies that build AI and companies that use it, from AI scale-ups to Fortune 1000 enterprises¹. If you sell an agent, read the indemnity clause in your customer contracts: Ollive lists indemnity obligations owed to enterprise customers among the exposures it plans to cover⁹.

How does AIUC-1 certification link to agent insurance?

AIUC publishes AIUC-1, a security, safety and reliability standard for AI agents. It audits companies against the standard and links insurance terms to the result².

ElevenLabs announced in February 2026 that it was the first to go live with an AIUC-1-backed policy, for its voice agents^{10, 11}. AIUC refers to its insurers without naming them, and no limits or terms were disclosed^{10, 12}.

Trade press reported in May 2026 that AIUC secured underwriting capacity from Beazley for a specialist AI liability product, but the only write-up we could read is an aggregator piece¹³. We found no mention of Beazley on AIUC's homepage or in its ElevenLabs update^{2, 10}. The AIUC-1 guide covers the standard in more detail.

Does your existing E&O or cyber policy cover an AI agent?

It might, but you cannot assume it. Policies written before generative AI usually say nothing about agents, and some insurers are moving to exclude AI outright.

Agent errors can also cause your own losses, such as the cost of pulling an agent offline. Beazley's AI Voluntary Shutdown Endorsement is for a business that suspends its own malfunctioning AI system to prevent further financial harm¹⁸. Relm's RESCAAI pays response and loss costs when an AI system you rely on goes wrong¹⁹.

If your agent depends on another company's model, an outage there becomes your problem. Coalition's AI page lists AI tool outages among the events its Active Cyber policy responds to¹⁶.

The coverage checker lists the clauses to look for, and does tech E&O cover AI goes through the E&O side.

What do insurers look at before covering an AI agent?

Evidence about the agent itself, not only about the company. The listings that describe their underwriting point to testing, certification or monitoring:

• Armilla's policies come with independent AI system certification and risk reporting¹.
• AIUC ties insurance terms to an audit against AIUC-1².
• Ollive presents itself as a runtime monitoring and evidence provider that supports underwriting⁷.
• Mayflower says it scores AI-specific risks such as bias, drift and hallucination²⁰.

Expect questions about what the agent can do without human sign-off, which tools and systems it can reach, how its actions are logged and how fast it can be switched off. These are general underwriting themes, not published criteria from any one insurer.

Contracts matter too. If you promise customers that an agent will meet a service level, check whether your policy excludes liability you take on by contract.

How much AI agent insurance can you buy?

Armilla and Vanguard AI are the agent listings that publish limits. Armilla's policy goes up to $25M per organization¹, and Vanguard AI states AI aggregate limits of $25M or more alongside $10M of cyber⁴.

AIUC, Corgi, Mayflower and Hadron, and Ollive publish no limit on the pages we read. No insurer we track publishes prices for agent cover.

AI Insurance Tracker does not sell insurance. A licensed broker can confirm whether a quote covers the specific actions your agent takes. For the wider market, see AI liability insurance and AI hallucination insurance.