Your AI Vendor Just Told Its Investors the Insurance May Not Be There
Picture a risk manager reading an SEC registration statement and stopping on one sentence. The company was Anthropic, disclosing risks ahead of a planned IPO. It wrote that "no single insurer, and potentially the insurance industry in the aggregate, may have sufficient capacity or expertise to underwrite insurance commensurate with such exposure" in a catastrophic AI-related incident.
It's not an insurer saying that. Insurers have been warning about AI exposure for months, and leaders have had time to discount it as self-interested caution from an industry that profits from charging more. This is different. The company building the product is telling its own investors the insurance market underneath it may not be able to absorb what happens if it fails at scale. Andrew Hill, global FINEX AI risk strategic leader at Willis, told Business Insurance that "it cannot be assumed that all AI exposures will be insurable."
Two numbers sit underneath the warning. According to Dow Jones reporting syndicated October 16, Aon estimates that more than 90% of AI risks covered by insurance fall into the "silent" category: present, but never explicitly addressed in the policy language (Dow Jones via Tiger Brokers). Munich Re's policy covering AI model errors tops out at $50 million. Hold those two side by side.
Three days after that report, W. R. Berkley's CEO made the carrier side of this plain on the company's October 19 earnings call. Rob Berkley said the company is working "very diligently to make sure the exposure... is identified, understood, controlled, and we're getting paid for it," and acknowledged applying exclusionary language "when we feel is appropriate" (W. R. Berkley Q3 2026 call transcript). Read that quote twice. "We're getting paid for it" means the exclusion isn't a courtesy. It's pricing and capital discipline.
Coverage for AI risk may exist. Most of what exists is silence, and silence leaves the answer to whoever reads the policy after a claim. Direction belongs to whoever calls the broker first. The risk manager who asks before renewal gets an answer. The one who waits finds out during a claim.
Here is the audit. Three questions for every renewal call this quarter:
- Has your carrier filed or attached an AI exclusion endorsement on any line you carry?
- Would your policy respond to a claim that arises from an AI failure, or only to language that sounds like it would?
- What do your AI vendor contracts promise in indemnification, and could that vendor's balance sheet back the promise in a claim large enough to need insurance in the first place?
Read the full research behind this story, plus this week's senior housing and governance flagships, in the LITT Briefing on Substack.