Insurers Are Dropping AI Coverage. Most SMEs Haven't Noticed.

Since January this year, insurers globally have had access to new endorsement forms that explicitly exclude AI-related claims from commercial general liability and professional indemnity policies. Not buried in dense legalese — printed in plain English on standardised forms designed to be stapled to your next renewal.

Most Australian SMEs haven't noticed. Tego CEO Eric Lowenstein, speaking at the UAC Market Exchange in Sydney, noted that widespread AI exclusions aren't yet standard across Australia — but overseas insurers are already adding them. Once exclusions become standardised, he cautioned, they spread fast through underwriting guidelines and renewal templates. The same trajectory as cyber exclusions a decade ago.

Before 2026, most liability policies were "silent" on AI — the technology wasn't explicitly included or excluded. If an AI tool contributed to an error that led to a claim, whether your insurer would pay was an open question.

That ambiguity ended on 1 January 2026 when Verisk released endorsement Form CG 40 47, a commercial general liability exclusion for bodily injury, property damage, and personal or advertising injury arising from generative AI. The form defines generative AI broadly: any machine-based learning system or model trained on data with the ability to create content or responses, including text, images, audio, video or code.

Separately, W.R. Berkley introduced Form PC 51380 — an absolute AI exclusion for Directors & Officers, Errors & Omissions, and Fiduciary Liability policies. The coverage lines that directly protect professional services firms. The Berkley exclusion covers any claim "based upon, arising out of, or attributable to" the actual or alleged use of artificial intelligence by any person or entity.

These are US forms. But as Insurance Business Australia reported, the Australian market is expected to follow the same path — from broad coverage to explicit exclusions to dedicated standalone AI products. Australian and London-market underwriters are already mirroring these clauses in their own policy wordings.

The sting is in the scope. These exclusions don't just cover the AI tools your team actively chose to use. They cover embedded AI — the features your software vendors added without asking.

Xero's AI categorisation engine. MYOB's AI-powered bank feeds. Your practice management system's smart scheduling. The dictation app your receptionist uses. If any of these technologies contributed to a circumstance that produced a claim, a broad AI exclusion could deny coverage — even if AI was incidental to the error.

For accountants and lawyers, professional indemnity insurance isn't optional. It's a licensing requirement. CPA Australia mandates minimum cover ranging from $2 million to $75 million depending on services and fees. Law societies in NSW, Victoria, and Western Australia require equivalent cover. If your PI policy now excludes AI-related claims, and your practice uses AI in any form, you have a material gap between your licensing obligation and your actual coverage.

The insurance implications of shadow AI just got sharper too. If your insurer can demonstrate that unapproved AI tools contributed to a claim, the exclusion applies whether you knew about the usage or not.

Here's the nuance that matters: coverage isn't necessarily gone. The new exclusions are often triggered not by AI use itself, but by AI use without verifiable governance.

Underwriters are now asking for documented artifacts. An AI tool registry listing every application capable of producing or informing client output — including browser extensions and productivity integrations. Contemporaneous review logs showing what a human reviewer changed, verified, or queried. Session-level records generated by the workflow, not reconstructed from memory after a claim.

The firms that entered 2026 with documented governance frameworks found they could negotiate affirmative AI coverage. The rest are discovering that "we always check the output" doesn't satisfy an underwriter's evidence threshold.

First, ask your broker a direct question: does your current policy contain an AI exclusion, endorsement, or carve-out? If they can't answer without checking, that's your answer — and your signal to read the policy yourself.

Second, build an AI register. Every tool your team uses that involves any form of machine learning or generative AI — including features embedded in vendor platforms like Xero, MYOB, or your practice management software. Browser extensions, ChatGPT logins, transcription apps — all of it.

Third, start a review log. When AI assists with client work — a draft letter, a transaction categorisation, a schedule optimisation — document what the human reviewer changed or verified. Timestamped. This is the edit delta that underwriters look for: evidence that oversight was substantive, not ceremonial.

None of this requires new technology. A shared spreadsheet and a fifteen-minute team brief will get most small firms to a defensible position before their next renewal conversation.

Key takeaways

Not sure whether your AI use is putting your coverage at risk?

A short call can map your current AI exposure against your policy and help you prepare for your next renewal conversation. Book a call.