Everyone's Worried About AI Running Out of Power. What Happens If It Runs Out of Insurance?

Insurers are writing generative AI exclusions into policies right as companies are getting ready to give AI systems more autonomy. Here's what I could verify. In January 2026, ISO, the group behind the standard general liability forms, made three generative AI exclusion endorsements available to carriers. Chubb, Berkshire Hathaway, Travelers, W.R. Berkley and AIG have filed to adopt them or something similar. One review of state filings found 2,369 generative AI exclusion records across 49 states. Berkley seems to have gone furthest, with an absolute AI exclusion for its D&O, E&O and fiduciary liability products. The ISO forms themselves are for general liability. What I can't tell you is how many renewals have actually dropped AI coverage. The filings show the exclusions exist and carriers can use them, not how often they do. The correlation problem is the part I find most interesting. Insurers deal with correlated risk all the time, including hurricanes and cyberattacks. But can they measure and cap AI exposure across their policyholders? If hundreds of customers run on the same underlying model, one failure could set off a lot of claims at once, and there's very little history to price it with. The scale is early too. In October 2025, the Financial Times reported, citing people familiar with the matter, that OpenAI had about $300 million of coverage for emerging AI risks, a small fraction of what its lawsuits could cost. Those are unnamed sources, and the coverage concerns the labs' own legal exposure, not enterprise deployments. Companies don't need insurance to deploy AI. They can keep the risk, push vendors for indemnities, or limit what the system can do on its own. I haven't found a case of a company shelving an AI deployment because it couldn't get coverage. Still, insurers are preparing to limit their exposure just as businesses are preparing to give AI systems greater responsibility. If businesses end up keeping more of the liability themselves, does that change how much autonomy they're willing to give these systems, and how much they're willing to pay for them?

CryptoAlerta — análise de criptomoedas e mercado em tempo real