When a client asks their broker which insurance company covers AI liability, most brokers have to admit they don't really know.
That's not a knock on brokers. The market for AI-specific coverage is genuinely new, genuinely fragmented, and genuinely inconsistent. A handful of carriers have built products for it. Most standard commercial insurers haven't. And the carriers who have built products often cover one type of AI risk and exclude another.
What follows is the most complete breakdown of which carriers are actually writing AI liability coverage in 2026, what they cover, and what they don't. No single independent comparison like this exists anywhere. Most brokers find out by submitting and seeing who responds.
The dividing line that changes everything
Before getting into specific carriers, there's one distinction that determines which market you belong to.
Are you a company that uses AI tools? Or is AI your product?
Most carriers writing AI liability coverage will cover AI users, meaning businesses that have added AI tools into their operations. A manufacturer using an AI procurement agent. An accounting firm running an AI tax assistant. A property manager with an AI chatbot handling tenant questions. These companies are AI users, and there are more options for them.
Companies whose primary product is AI, meaning they're building and selling the AI itself, face a much harder market. The liability profile is fundamentally different. When your product fails, your client absorbs the loss, and the exposure flows back to you at scale. Most of the newer specialty carriers cover AI users and exclude companies whose primary product is AI, either explicitly in the policy form or through how the underwriting questionnaire is built.
If you're an AI builder, you're not without options. But you need to know going in that the standard commercial E&O market isn't your market, and the carriers below have very different appetites for builders versus users.
Armilla
Armilla operates as both an underwriter and an AI risk scoring platform. Their approach is unusual in that they actually assess the AI system before binding coverage. Underwriting is based on a specific risk profile, not a general description of what the business does.
Armilla focuses on what they call AI performance risk, meaning claims that arise when an AI system's output causes financial harm. They write professional liability forms designed around AI-specific exposure, including coverage for errors in AI recommendations, failures in AI-generated analysis, and harm caused by AI decisions that turn out to be wrong.
Their market skews toward AI users in professional contexts: financial services, healthcare-adjacent businesses, legal tech. They're also one of the few carriers that will engage with AI builders under the right conditions, though the risk scoring requirement makes the process more intensive.
If you're running AI tools in a professional services context and your existing E&O carrier is adding exclusions, Armilla is worth a submission. They're not the cheapest option, but they're building products that actually match the exposure.
Testudo (Lloyd's)
Testudo writes through Lloyd's of London, which matters as context. Lloyd's syndicates have historically been the market where coverage gets invented for risks that traditional admitted carriers won't touch. Asbestos liability, cyber insurance in the 1990s, event cancellation. Lloyd's figures it out before the standard market does.
Testudo's AI liability form covers professional liability and errors and omissions specifically for AI-related outputs. They're built around third-party liability: the situation where your AI agent's output harms a client or a third party and someone sues you over it.
They cover both AI users and, selectively, AI builders. The builder side requires more underwriting information and tends to come with higher retentions. If you've been declined by an admitted carrier, a Lloyd's market is often the next call worth making.
Access runs through a surplus lines broker. If your standard broker hasn't submitted to a surplus lines market for your AI risk, that's a question worth asking directly.
Relm
Relm came up in the emerging tech market covering crypto and digital asset companies and expanded into AI liability as the exposure became clearer. They're now one of the more accessible options for companies that would struggle in a standard admitted market.
Relm's focus is tech companies and startups, including AI builders. They're less oriented around traditional professional services and more oriented around the tech company profile: software companies, AI platforms, data companies. If your risk involves IP exposure, data liability, or AI output that touches multiple downstream clients, Relm belongs in a marketing submission.
They write primary and excess layers. That matters if you're a startup with a contractual minimum that doesn't reach the limits your investors or enterprise contracts require.
Mayflower Specialty
Mayflower Specialty operates in the excess and surplus lines space, which means they're built for risks that don't fit standard market appetites. AI liability is exactly the category they exist for.
Their AI-related products focus on technology E&O and cyber, with specific endorsements available for AI output liability. They're not a dedicated AI insurer the way Armilla is, but they're a flexible market that can structure a program around a risk profile that doesn't fit anywhere else.
For Illinois businesses that have been declined by their existing E&O carrier because of AI-related activity, Mayflower is a realistic next step. They're also used to working with accounts that need manuscript endorsements, meaning customized policy language rather than a standard form off the shelf.
Munich Re / HSB
Munich Re is the reinsurance giant behind a significant portion of AI risk capacity in the market right now. HSB (Hartford Steam Boiler) is their U.S. operating subsidiary, and they've built a direct AI-related product line focused on equipment breakdown and technology risks.
HSB's AI coverage focuses on the physical and operational side of AI failure: systems that control equipment, manufacturing automation, infrastructure. If your AI risk involves a physical process, a factory floor, a building control system, a commercial fleet, their product line is worth exploring.
They're also a reinsurance backstop for several of the newer specialty markets. When Armilla or Relm can write your risk, Munich Re's capacity is often part of why they can do it at scale.
Corgi
Corgi offers usage-based AI liability coverage. Rather than a flat annual premium for a described program, their model is designed to track actual AI usage and price accordingly.
This is still emerging as a structure in the U.S. market. It's potentially well-suited for companies whose AI usage is variable, startups that haven't hit full deployment scale, or companies that use AI intensively in some quarters and minimally in others. The actuarial logic of usage-based pricing for AI makes sense. But the policy mechanics are newer, and it's worth reading the form carefully before binding to understand how "usage" is defined and what happens if your usage spikes unexpectedly.
Where traditional carriers actually stand
If you have an existing commercial program with a standard admitted carrier, your AI coverage depends heavily on when your policies last renewed and whether your carrier adopted the new ISO exclusions.
ISO filed AI-specific exclusions in January 2026, including endorsements CG 40 47 and CG 40 48 for general liability policies. Carriers aren't required to adopt ISO forms, but many standard commercial insurers follow ISO closely. If your GL renewed after those filings and your carrier adopted them, AI-generated outputs may now be excluded from your general liability coverage. Some major carriers have filed AI exclusions with state regulators including Illinois. Others haven't moved yet.
Professional liability forms have been adding AI exclusions independently. Some carriers started doing this in 2024. Others haven't touched their forms. Whether your current E&O covers AI-generated professional errors isn't something you can assume either way. It requires looking at the actual endorsement schedule of your current policy, not the summary page.
Cyber policies are the third piece. Many cyber forms include some AI-related coverage, particularly around AI-enabled social engineering and data liability connected to AI tools. But cyber coverage wasn't designed to be the primary AI liability product. Relying on cyber to catch what GL and E&O miss is a gap, not a strategy.
And the standard admitted carriers who haven't added exclusions yet? That's a temporary situation. The market is actively evolving, and comfortable positions from 2025 are getting revisited as carriers learn what AI claims actually look like.
What Illinois businesses should actually do
Getting AI liability coverage right in 2026 isn't a one-carrier transaction. It's a program-level question.
For most Illinois businesses using AI tools but not building them, the path starts here: review your existing E&O and GL forms for AI exclusions, understand exactly what's in and what's out, and decide whether the gaps are material enough to fill with a specialty product. Some businesses will find their existing coverage is adequate. Others will find they have a real gap between what their GL covers and what their actual AI exposure looks like.
For companies building AI products or running AI in high-stakes professional contexts, the standard admitted market probably isn't going to get you there. The specialty markets above are where this coverage actually lives. A broker who knows those markets is worth more than a lower-cost one who doesn't.
The market is also moving fast. Carriers that declined AI submissions in 2024 are looking at them now. Carriers that were comfortable writing AI risk in 2025 have started adding exclusions as they see what claims look like. A program that was right twelve months ago may have real gaps today.
If you're an Illinois business with questions about where your AI liability actually stands, reach out to Jack Ray directly. He handles commercial accounts across the state and can review your current program against the markets writing this risk now.
Email: jray@lakeshoreriskadvisors.com