Korva is a scroll of the internet’s best ideas.
AI labs outgrew their insurance
Mounting incidents
The incidents are already piling up: rogue agents hacking into websites, chatbots allegedly encouraging people into suicide, and AI hallucinations at law firms. Where the liability for any of it falls is unclear, and how much existing policies would cover has not yet been tested. Insurers are weighing whether to exclude some of these risks outright.
Why scale worries insurers
Scale is what worries them. Only a few labs exist, so the risk is concentrated, and a hack that took a big bank down for a long stretch could do enormous economic damage. Nobody is willing to underwrite what insurers call a tail risk, and a civilization-ending event is one that no one will write a policy for.
The market responds
Startup insurers now focus only on AI risk and write policies aimed squarely at it, and the big insurers are studying the market too. The largest labs have deep pockets and can cover some liabilities themselves. With large lab IPOs coming, regulators and lawmakers are questioning the gaps, because an insurer's willingness to underwrite a risk shows that an outside company has looked at it and accepted it.
Why won't insurers cover tail risks?
Insurance works by pooling many modest, mostly independent losses so the premiums of the many pay for the few. A tail risk is a rare event so large that it could hit many policyholders at once, which breaks that pooling. Actuaries also price policies from historical loss data, and a new technology has very little of it. Reinsurers, the insurers of insurers, have long written exclusions for correlated catastrophes such as war and nuclear accidents for the same reason.
How did cyber insurance handle a similar gap?
Cyber insurance grew up amid doubt about how much a single attack could cost. Insurers worried about what they called silent cyber, meaning older policies that never mentioned digital losses but might still be forced to pay for them. In 2019, Lloyd's of London told its syndicates to state clearly whether each policy covers cyber losses. Many insurers have since added explicit exclusions or separate cyber policies.
Why do IPOs put insurance gaps in focus?
A company going public must describe its material risks to investors in filings with the securities regulator. In the United States, those appear in the risk factors section of the prospectus. Large unresolved liabilities can count as material, and the banks underwriting the offering and the auditors examine them closely. Public companies also commonly buy directors and officers liability insurance, which covers claims against executives over how they ran the company.
How can a company cover its own liabilities?
Large companies sometimes form a captive insurer, a subsidiary that exists only to insure the parent's risks. A captive lets the parent keep premiums in the family and retain risks it believes it understands better than the open market does. The captive can also buy reinsurance to pass the largest losses to outside firms. Bermuda and Vermont are two well-known places where captives are registered.
More from WSJ What’s News
