Skip to Content

AI Liability Risk for Technology Companies

Most technology companies didn't decide to adopt AI in a risk meeting. They decided in a product meeting, or a board session, or because a competitor moved first. The insurance question came later and for plenty of businesses, it still hasn't really come up. 

It's worth bringing forward. Of the five risk categories we scored in the Forerunner Report — Technology, AI and automation error ranked last. Not on likelihood or impact but on insurability. 

Scored by Knightcorp’s placement team, AI and automation error scores 3.5 out of 10 on our Forerunner Index. Cloud misconfiguration sits at 6. IP infringement at 7. (Higher scores signal better insurability potential , a general indicator of factors insurers weigh, such as placement, pricing and underwriting conditions.  

Specific outcomes always depend on the insurer, market appetite and your risk profile.   

The gap between 3.5 and 7 isn't a technical curiosity. It's a commercial constraint on the capability most technology companies are betting their growth on. 

The Reason Isn't What Most People Assume

It isn't that insurers think AI is uniquely dangerous. It's that many policies still don't say anything about it at all. 

The industry has a name for this now: silent AI. Wording that neither excludes AI-related risk nor affirmatively covers it, the same pattern the market moved through with “silent cyber” a decade ago, except on a far more compressed timeline and with a much more fragmented response. ² 
That gap isn't surprising given the pace of adoption. In KPMG's 2026 Keeping Us Up at Night survey, AI-related issues overtook inflation to become the top concern for Australian business leaders — a sign that rapid deployment is moving well ahead of the wording built to insure it. 

Two Insurers, Same Risk, Opposite Answers

Here's what makes this a live commercial variable rather than a future problem. The market isn't moving in one direction. It's splitting. 

Some Australian insurers are adding affirmative AI cover, wording that states explicitly how the policy responds when AI is involved, while others are writing exclusions. ⁴ 

Two technology companies with near-identical AI exposure can end up with materially different coverage outcomes, shaped largely by which insurer they renewed with and what wording was in front of them at the time. That's not a risk-management problem. It's a placement problem and it's live at renewal now. 

The Exposure Most Companies Have Backwards

There's a common assumption that building AI in-house carries more liability than buying it.  In practice, the reverse may be true. 

When a business deploys a third-party AI tool, it may inherit risks it cannot fully see: the model’s limitations, the data and assumptions behind it, and the way it produces recommendations or decisions. Yet if that tool affects customers, employees or other stakeholders, attention is likely to turn first to the business that selected it and put it into use. 

Recent US matters illustrate the distinction. 

  • In Walters v OpenAI, a radio host sued OpenAI after ChatGPT generated a false statement about him. The Georgia court ruled in OpenAI’s favour, finding that the chatbot output was not legally defamatory in the circumstances. The decision reflected, among other factors, ChatGPT’s warnings that its responses may be inaccurate, the context in which the output was generated, and the absence of demonstrated harm.⁵ 
  • The more practical exposure can sit with the organisation using AI. I In July 2025, Earnest Operations, a student-loan provider, agreed to a US$2.5 million settlement with the Massachusetts Attorney General through an Assurance of Discontinuance. The AG alleged its AI underwriting models could cause disparate harm to certain applicants, and the settlement also covered other practices. Earnest denied the allegations. —but agreed to ongoing governance, testing and compliance measures.⁶ 

The lesson is not that AI vendors escape scrutiny. It is that companies cannot outsource accountability simply by buying the technology. If an AI tool informs customer-facing decisions, eligibility, pricing, communications or advice, the business deploying it needs to understand the risk, test the outcomes and retain clear human oversight. 

Different jurisdictions, different legal theories — but the direction of travel is clear: businesses cannot assume that buying AI transfers accountability. The organisation that deploys the tool is often the most visible party, and the one expected to explain how it governed its use. 

Where the Matrix Puts It

The Forerunner Report's own Risk Prioritisation Matrix shows how AI/Automation Error compares to the other four risk categories are scored in the report. 
If we had to explain the problem in just six words, its this“emerging claims unclear policy fit." It isn't that AI claims are rare. It's that when they happen, nobody whether its insurer, broker, or business is fully certain which policy is meant to respond. 

Download the Forerunner Report - Technology

The Forerunner's Move

The businesses ahead of this aren't slowing AI adoption.  They are getting ahead of the insurance question. 

Before renewal, they are mapping where AI is being used, what could go wrong and whether their existing insurance programme responds — or is silent — on the risk. 

That means matching specific AI uses to the Covers most likely to be tested:  


A policy silent on AI isn't the same as a policy that covers it. And last year's wording may not be this year's. 

What Is AI Liability Risk for Technology Companies?

AI liability risk is the exposure that arises when AI systems make or shape decisions affecting customers, employees or counterparties, and it is unclear which insurance policy, if any, would respond to a resulting claim. 

It ranked lowest of the five risk categories in the Forerunner Report on insurability. This is largely a wording problem: many professional indemnity, cyber and D&O policies were written before AI-driven decisions became common, so they don’t address AI either way.

The Knightcorp Point of View

Insurability is not a compliance exercise. It's a measure of how easily a business can transfer a risk it has chosen to take, which makes it a growth issue, not a paperwork one. 

For AI, the question is not whether a policy exists. It is whether the insurance programme would respond when the business uses AI to make decisions, serve customers, handle information or deliver products and services. 

We use the full market, clear strategy and direct engagement to pursue the best available outcome for the client. On AI specifically, that means testing how a programme would actually respond to the decisions a business is already making, rather than confirming a policy exists. 

We turn assumption into evidence, and exposure into a plan. That clarity, ahead of the market, is the forerunner's edge. 

Read the Forerunner Report - Technology

Frequently Asked Questions

  1. Is AI risk covered under my existing professional indemnity policy? 

Often it's neither clearly covered nor clearly excluded — the industry term is “silent AI.” Whether a policy responds depends on its specific wording, and many policies drafted before autonomous AI became operational don't address the question at all. 

  1. What does “silent AI” mean? 

Policy wording that neither expressly excludes AI-related risk nor affirmatively covers it. It mirrors the “silent cyber” phase the market moved through before cyber became a separately-worded exposure. 

  1. Are insurers adding AI cover or taking it away? 

Both, concurrently. Some Australian insurers are introducing affirmative AI wording; others are writing AI exclusions. That divergence is why placement matters more than usual right now. 

  1. Am I more exposed if I build AI myself or use a third-party tool? 

In our view, third-party deployment often carries the greater practical exposure — you inherit the model's risk without visibility into how it was built or trained, attention and potential liability may fall on the business deploying the tool, rather than only the developer.  

  1. Does AI liability only affect companies whose product is AI? 

No. It affects any business using AI to make or shape decisions about people — hiring, pricing, credit, onboarding, eligibility, customer service. The exposure follows the decision, not the product category. 

  1. What does “unclear policy fit” actually mean for an AI-related claim? 

It means that when an AI-related loss occurs, there's often no obvious answer as to which policy — professional indemnity, cyber, or D&O — is meant to respond, because AI claims don't sit neatly inside any one of them. That ambiguity, not the frequency of claims, is what the Forerunner Matrix is flagging. 

  1. Why is AI harder to insure than cyber or IP risk? 

Because many policies were priced and worded for risks with an established claims history and a clear policy home. AI hasn't had either for long enough — its potential impact is high, but the market hasn't yet settled on which policy is meant to carry it. 

  1. What should a technology company check before its next renewal? 

Whether the policy mentions AI at all, and if so, in which direction — and whether the wording has changed since the last renewal.