OpenAI, Google, Meta, Anthropic lack insurance against catastrophic AI risk

The companies racing to build the most powerful AI systems on the planet have something in common. None of them appear to be insured against the disasters they keep warning about.

When asked, none of the AI companies, a group that includes OpenAI, Google, Meta, and Anthropic, pointed to insurance covering catastrophic risk. The firms that write the scariest risk disclosures in tech are, in effect, carrying that risk themselves.

Nobody wants to underwrite the apocalypse

Catastrophic AI risk breaks the standard insurance model. The scenarios in question involve mass harm or broad societal disruption, losses so large that no insurer’s balance sheet could plausibly absorb them. Underwriting for these severe, low-probability, high-impact outcomes is largely missing from the market, according to the research findings.

That leaves AI developers with a short menu of options. They can buy narrow policies with limited coverage, insure themselves, or lean on investor capital to absorb shocks.

OpenAI has minimal insurance coverage for emerging AI risks, with figures around $300 million circulating, though that number is contested. Larger AI firms are increasingly turning to self-insured arrangements through captive insurance companies, where the company becomes its own insurance provider.

Anthropic put the warnings in writing

Anthropic’s IPO prospectus, released in late September 2026, warns of “catastrophic or existential risks to humanity” from advanced AI. The company devoted approximately 80 of the document’s 261 pages to risks tied to advanced AI, including various self-preserving behaviors that could threaten humanity.

Anthropic’s potential valuation has been discussed at as much as $2 trillion. The company is asking public investors to fund it while also documenting, at length, the ways its technology could cause harm that no insurer will cover.

Regulators may end up writing the policy

Some in the insurance community want mandatory liability insurance tailored to existential risks from AI development. Yoshua Bengio, the Turing Prize winning AI researcher, has advocated for AI firms to be required to carry liability insurance covering catastrophic risks, drawing on the model from nuclear power, where regulators have long required specific insurance arrangements.

One proposal is a mutual insurance company for frontier AI, where developers would pool risk among themselves. Supporters suggest it could also enforce safety standards as a condition of membership and guarantee liability coverage where the traditional market will not.

Catastrophe bonds are another option being floated. These instruments let investors take on specific disaster risks in exchange for returns, and they could target scenarios like catastrophic failures involving critical infrastructure.

What this means for investors and the industry

Without meaningful coverage, potential large-scale liability claims sit directly on the companies themselves. If mandatory insurance requirements arrive, AI companies would face new costs, and the structure of those rules would matter a great deal.

How investors price Anthropic at its public market debut — a company that spends roughly 80 pages describing its own potential dangers — will signal how seriously markets take the uninsured tail risk.