„Insurers struggle to price AI-caused damages, so they narrow coverage with exclusions, leaving the liability gap of AI agents largely uncovered.”
Counter-narrative
Beazley, CFC, Munich Re, and new players (Harbor) offer explicit AI coverage. Some cyber insurers clarify existing terms without introducing new exclusions, while some specialty insurers resist exclusions. The CEO of Marsh says broad AI exclusions could cause uncertainty.
Carriers
Insurance Journal and the insurance trade press, Chubb and Marsh executives, Beazley, CFC, Munich Re, Financial Times, New York Times, Anthropic (securities filing)
Analysis
The narrative holds that AI liability is the insurance market's new blind spot: silent risk is carried by traditional policies until exclusions push it out. It is fed by AI exclusions in D&O and E&O policies, broker warnings, and Anthropic's own risk statement. It opens a new product market for insurers and specialized startups (Harbor), and creates demand for advice among brokers. Its weak point is that without claims statistics, both exclusions and coverage rest on estimates, and the market is divided. Here the liability question meets the rogue-agents narrative and, through Anthropic's stock-market filing, the lab-IPO theme.
Evidence: moderate — Reports on changes to policy terms and on new products (Beazley, CFC, Munich Re aiSure, Harbor, Crosby's policy) support it. There is also a mention of an incident caused by a runaway AI and of reported losses. However, loss statistics and a large paid-out AI claim do not appear in the stories, some of the reports on exclusions are about contemplated moves, and the expected scale of losses is a forecast.
What's at stake: If AI damage is not insured, the risk remains with the customer and on the balance sheet.
Narrative sub-threads
The narrative splits into 4 branches; a story may belong to several branches, and some may not appear in the main narrative's list.
AI exclusions in liability insuranceliability insurance
4.3
Runaway agents in cyber insurancecyber insurance
0.9
Standalone AI hedges and specialistsinsurance coverage
no measurable search
0
AI harm at the limit of insurer capacityinsurance claims processing
no measurable search
0
World
Topic and search term
Weekly
Average
Insurers cannot price AI risk eitherai insurance
12.3
AI exclusions in liability insuranceliability insurance
15.8
Runaway agents in cyber insurancecyber insurance
3
Standalone AI hedges and specialistsinsurance coverage
12.7
AI harm at the limit of insurer capacitycatastrophe insurance
0
Google Trends relative search index over the period studied, weekly average. A narrative and its sub-threads appear in a single comparison: 100 is the group's most-searched day, and the figures can only be compared within the group. The search term was chosen by machine analysis; it measures how much the term is searched, not how widespread the narrative is.
There is hidden AI risk in the existing portfolio. For each policy, it must be determined whether AI risk is covered, excluded or expressly covered, and this must be recorded in clear wording.
AI exclusions are appearing in corporate D&O and E&O policies, while cyber clauses are being refined for AI-related damages. Upon renewal, specific inquiries must be made regarding damages caused by AI agents.
The terms of liability insurance for AI agents are still taking shape, but such policies are already available: the AI law firm Crosby purchased one. It is worth assessing the availability of explicit AI coverage early.
Hungary
Colonnade Insurance analyzed how AI-driven cyberattacks are assessed in cyber insurance. The Hungarian professional press raised the issue that corporate insurance does not necessarily provide a clear answer on who pays if AI makes a mistake. The MABISZ review wrote about Hungarian insurers' own AI usage. There is no news on an independent Hungarian AI liability product; it is worth monitoring the terms and conditions modifications by MABISZ and major insurers.