Insurers are hesitant to offer broad corporate liability coverage for AI-related harm, according to a RAND report covered by The Register. The hesitation is not driven by apocalyptic fears about superintelligent machines, but by a more mundane problem: there is almost no historical claims data to calculate the likelihood and cost of AI-caused incidents.
RAND argues that the industry needs better data on real-world AI failures and their consequences before underwriters can confidently price policies. Without that information, insurers either avoid the market or charge premiums that may not reflect actual risk.
The report distinguishes between ordinary corporate liability concerns and extreme scenarios, saying the latter are not the main obstacle. The Register's coverage notes that RAND wants a more empirical approach to machine-made mishaps, treating them as a data problem rather than a speculative one.