AI Losses Surge, but Insurance Coverage Remains Unclear as One Flaw Could Trigger Hundreds of Millions in Damage
November 18, 2026
AI-related losses are rising, but Allianz Commercial warns that reported claims understate businesses’ real exposure: critical AI incidents increased by more than 150% over two years, and a single loss could reach hundreds of millions of dollars.
The concern is that AI is spreading rapidly across industries and countries, often through shared models, training data, cloud infrastructure and data centers, so one flaw could ripple across many organizations.
A study published in China Finance Review International argues that AI risk is not just cyber risk, because internal system or governance failures can cause harm even when no system has been hacked.
AI systems can generate convincing but incorrect outputs because they rely on probabilities rather than distinguishing truth from falsehood, and shared dependencies can turn small faults into widespread damage.
The litigation landscape already spans intellectual property and copyright, privacy and unauthorized data use, inaccurate outputs, discrimination, algorithmic bias and environmental impacts; more than 100 intellectual-property lawsuits against AI developers had been filed by mid-2026.
Existing insurance claims include alleged failures in vehicle safety systems and shareholder lawsuits alleging misleading disclosures or omissions about companies’ AI use.
For businesses, the coverage picture remains murky: a 2026 RAND Corporation report found some insurers affirmatively cover AI losses, others apply broad exclusions, and many policies are silent, leaving outcomes dependent on how a claim is characterized and which policy line is involved.
An AI-related loss could implicate cyber, technology errors and omissions, directors and officers, or employment practices liability coverage, so clients should not assume cyber insurance automatically covers AI exposure; AI is a tool, not a peril, and coverage depends on the loss.
Insurers are also tightening language: optional ISO generative-AI exclusions apply to commercial general liability forms, while W.R. Berkley has adopted an absolute AI exclusion across some D&O, errors and omissions, and fiduciary liability products.
Views on exclusions differ: Verisk’s Joe Lam says they can give underwriters confidence to accept other risks, while Allianz’s Melissa Hill warns that businesses could lose coverage when AI is involved in an operational shutdown; Alana McMullin expects D&O and E&O disputes to be early tests, and no leading court case has settled the issue.
The study says clearly attributable operational errors, specific liability claims and AI-enabled cyber incidents may be insurable, while systemic failures, unclear accountability and extreme losses with little historical precedent are much harder to cover.
To make coverage workable, the study recommends narrowly defined policies, conservative limits, scenario-based pricing and underwriting that examines model testing, human oversight and governance; clearer liability rules and transparency requirements could also help.
Summary based on 16 sources
Get a daily email with more AI stories
Sources

Insurance Business • Nov 19, 2026
AI losses are hiding inside existing policies, Allianz Commercial warns
Mirage News • Nov 20, 2026
Why AI Risks Are Harder to Insure Than Cyber Risks
BIOENGINEER.ORG • Nov 20, 2026
Study Finds Internal Flaws Make Artificial Intelligence Risks Harder to
DagangNews - Berita Bisnes Anda • Nov 19, 2026
Allianz: Reported AI losses understate the true risk and are only the tip of the iceberg| DagangNews - Berita Bisnes Anda