01
Non-determinism
No defect to inspect. Only a failure rate.
01 · The carve-out
The exclusion is official.
ISO filed generative-AI exclusions for general liability. Chubb, Travelers, W.R. Berkley and Berkshire Hathaway have filed to adopt them.
An industry writes exclusion forms when it can't price the risk.
CG 40 47
Exclusion: Generative Artificial Intelligence
This insurance does not apply to "bodily injury", "property damage" or "personal and advertising injury" arising out of generative artificial intelligence…
Related filings: CG 40 48 · CG 35 08
02 · The structure of the problem
Underwriting breaks seven ways.
02
Non-stationarity
The system changes under the policy. Month six is not day one.
03
No loss history
No claims data. Histories go stale faster than they build.
04
Self-reported evidence
Fire had inspectors. Agents have the insured's word.
05
Correlated failure
Thousands of insureds. A handful of models. No diversification.
06
Unmapped severity
Most errors cost nothing. A few cost everything. Nobody knows which.
07
Moving law
The EU wrote a liability directive, then withdrew it. Rules churn faster than the tail.
Any one would make a risk hard. This one has all seven.
03 · The aggregation problem
One model fails.
Every book bleeds.
Insurance pools independent risks. This risk isn't. Most agents sit on the same few foundation models. One bad update hits every book at once. A catastrophe line, no hurricane map.
Reinsurers can't model it, so they cap capacity. The largest standalone limit on offer: $25M per insured.
Smaller than the single Arup loss.
04 · The gap, measured
A handful
of standalone AI-liability products worldwide
$25M
the largest standalone limit on offer, per insured
$25.6M
one company, one deepfake, one day
$1.5B
the largest copyright settlement on record
40%
of enterprise apps will feature task-specific agents by end of 2026
€35M / 7%
EU AI Act penalty ceiling
Deloitte: $4.8B in premiums by 2032. Premiums are the priced fraction. The rest sits on balance sheets.
05 · The precedent
"Uninsurable" has been wrong before.
1909
Railroad bonds
Bought on the issuer's word. Moody's graded them. Markets moved on the grades.
1996
Driving behavior
Priced by proxy. Progressive put an instrument in the car.
1997
Cyber
AIG wrote it with zero actuarial data. It became the industry's fastest-growing line.
Insurability arrives with one thing: independent, continuous measurement.
An AI agent is a driver without telematics.
The point
Capital showed up.
Measurement didn't.
That's what Klira is building. If you underwrite, reinsure or broker this risk, let's compare notes.
Or get early access by email