AI risk used to sit mostly with the companies that build AI. That’s changing. Today, a business can be exposed because it makes AI, because it uses AI in its own work, or simply because it depends on someone else’s AI to keep running.
Most liability policies say nothing about AI. That silence is what’s worrying brokers, and some carriers are now responding with exclusions. Relm takes the opposite approach: affirmative AI cover, written to say what it covers.
“At the end of the day, insurance is all about trust.”
Claire Davey, Relm Insurance
At Relm, we’ve built our AI product suite around those three ways a business meets AI: building it, using it, or relying on it. Here’s how it fits.
- PONTAAI sits excess of your client’s existing liability program, whatever the line: general liability, D&O, professional indemnity, tech E&O, and more. If the underlying policy is silent on AI, excludes it, or erodes, PONTAAI can respond. Your client keeps the tower they have and gets affirmative AI cover without disrupting existing relationships.
- For companies that build AI or sell AI-enabled products and services, NOVAAI is the primary policy: cyber and tech E&O built for AI, with affirmative cover for AI regulatory fines, AI-generated media, and AI product liability.
- For businesses that rely on third-party AI, RESCAAI covers the first-party fallout when it fails: business interruption, reputational harm, product recall, and incident response.
[CONFIRM WITH SHANE BEFORE PUBLISHING: the A- rating and paper for the AI products. If it can’t be confirmed, delete this line, the heading, image, and paragraph below. If it’s confirmed, delete just this line and upload the image as AI-A-Rated-Paper.png.]
Specialist AI Cover on A- Rated Paper
Our AI products are written on A- rated paper, so you don’t have to choose between specialist expertise and strong security.
Five Changes Worth Knowing About
Our latest research looks at five changes, grouped by how a business meets AI: building it, using it, or relying on it.
1. Reported AI harms are rising
The AI Incident Database, which collects documented cases where AI systems have caused or nearly caused harm, recorded 362 incidents in 2025, up from 233 in 2024. A second tracker, run by the OECD, logged a monthly peak of 435 incidents in January 2026 (Stanford HAI AI Index 2026). For a company that builds AI, each failure can turn into a claim. These are reported incidents, mostly drawn from news, so some of the rise may reflect more reporting, not only more harm.
2. AI laws are multiplying
US states passed 150 AI-related bills into law in 2025, up from fewer than 10 in 2020 (Stanford HAI AI Index 2026). In the EU, most of the AI Act’s transparency rules have applied since August 2, 2026, while the main rules for stand-alone high-risk AI systems have moved to December 2, 2027 (K&L Gates). Some rules apply to companies that build AI, others to those that use it. Not every AI-related law adds new duties, and US federal policy is pushing back on some state laws, so this picture may keep shifting.
3. AI now ranks second only to cyber
In the Allianz Risk Barometer’s ranking of business risks, AI jumped from #10 to #2 in a single year, behind only cyber (Allianz). Use is now widespread, too: 88% of respondents to a McKinsey survey say their organization uses AI in at least one business function (Stanford HAI AI Index 2026). Once AI is part of everyday work, the questions are about decisions: what AI is used for, what customers are told, and who checks the output.
4. Most policies are silent on AI. Some are excluding it.
Most liability policies say nothing about AI either way. In January 2026, Verisk released three optional ISO endorsements for use that let insurers exclude generative AI from general liability cover (IndependentAgent.com). Berkley has introduced a so-called “Absolute” AI exclusion for its D&O, E&O, and fiduciary liability products (Hunton Andrews Kurth). The forms are optional, but they’re being used: one review of state filings counted 2369 generative AI exclusion records in force across 49 states by mid-2026 (Insurance Business). But a policy that’s silent on AI leaves the answer unclear, and an endorsement added at renewal can settle it the wrong way, so it’s worth checking what was added to a client’s policies at the last renewal.
5. Firms depend heavily on their AI vendors
In an IBM survey of 1000 senior executives, 81% said a seven-day outage at their primary AI vendor would be severe or critical, and 71% said switching their primary AI vendor or model would be difficult (IBM Institute for Business Value). The survey leans toward large firms, and these are what executives expect, not measured outages. Still, for a business that relies on someone else’s AI, the risk isn’t only a claim from outside. It’s the AI stopping or being breached, and the business losing income as a result.
What This Doesn’t Show
This research doesn’t show rising AI insurance claims, a shortage of insurance capacity, or a gap in any client’s cover.
The Question Worth Asking
Which of your clients build AI, use it in their work, or depend on someone else’s, and has their cover been reviewed since that changed?
None of this means a client’s current cover falls short. But if a client builds, uses, or depends on AI in ways it didn’t at its last renewal, now’s a good time to have the conversation.
Why Relm
- Three dedicated solutions for providers and users of AI: NOVAAI, PONTAAI, and RESCAAI.
- Affirmative cover, not silence. Our wordings say what they cover.
- Broad appetite and deep expertise. We’ll consider risks many markets won’t, including AI companions, agentic AI, and generative AI content, subject to underwriting.
We’re also working on AI risk where adoption is moving fastest. With DIFC Innovation Hub, Relm convened the Emerging Risks Lab. Its first report, Scaling AI in banking: navigating a new era of risk and resilience (September 2026), draws on a workshop of 23 senior executives from banks, regulators, law firms, AI scale-ups, and technology providers across DIFC and the UAE. They raised 56 risk statements, which the Lab consolidated into 26 risks across six themes. Its conclusion: “Most of the risks discussed in this report are insurable today… However, too many insurers remain ambiguous regarding AI technologies and thus fail to provide boards and executives with the certainty that they are seeking when deploying such technologies.” Read the report.
In October, Relm exhibited at the Dubai AI Festival in Dubai. CEO Joseph Ziolkowski spoke on two main-stage panels, and we hosted a roundtable on the Emerging Risks Lab findings.
Get the Full Research Report
These five changes come from our new research report, Silent on AI. It sets out the data behind each trend, what it means for clients, and the questions worth raising, with every figure traced to its original source.
Download Silent on AI (PDF)
Does your firm have casualty, P&C, or financial lines teams? Introduce us. PONTAAI sits above the programs they already place, and it gives them a way to talk to clients about AI without replacing anything.
To talk through PONTAAI, NOVAAI, RESCAAI, or a specific client, contact our team:
This article is for general information only and isn’t legal, financial, or insurance advice. Coverage is subject to the terms, conditions, limitations, and exclusions of the applicable policy wording, and products may not be available in all jurisdictions.
