When Algorithms Sail Without Owners: Insurance Design for Agentic AI Harm

Abstract

As businesses increasingly use agentic AI systems, autonomous agents capable of carrying out complex, multi-step tasks without constant human supervision, we're facing a major crisis in insurability. Leading insurers like AIG and W.R. Berkley are moving to exclude AI-related liabilities from standard corporate insurance, and the few specialized AI insurance options that exist are still in their early stages and not widely available. The risks that agentic AI poses don't align with traditional insurance, which relies on identifying specific dangers, calculating probabilities, and covering accidental losses. Inspired by two centuries of maritime mutual insurance, particularly the Protection and Indemnity Clubs established in 1855, I propose the creation of industry-specific AI Mutual Risk Pools to address this gap. Compared with voluntary audits or conventional insurance, mutual pools bring together those exposed to similar risks, encouraging strong safety practices and ensuring victims are compensated. By examining today’s insurance market, new regulations, and the similarities between maritime and AI risks, I argue that mutual pooling could address the unique liabilities of autonomous systems.