Insurers are already tightening terms. W. R. Berkley has added exclusions for any use, deployment or development of AI across its D&O, E&O and fiduciary products. Broker Aon's analysis of more than 300 AI-related disputes found potential exposure across crime, IP, cyber and technology E&O lines. Cyber insurers including MSIG, QBE and Beazley are revising policy wording for autonomous systems. RAND reports that fast enterprise AI adoption is outpacing a fragmented insurance market, leaving companies exposed to losses from hallucinations, deepfakes and discriminatory outputs. Courts have not yet set clear precedents for how existing liability rules apply to AI. Underwriters cite that gap as a central source of uncertainty. How much of the estimated exposure will fall on insurers versus the labs, and whether a government backstop or mandatory pool materializes, remains unresolved.
The story matters now because the incidents, the scale of potential claims and the market response are converging. Private efforts are also emerging. The Artificial Intelligence Underwriting Council has $55 million in backing to develop AI auditing standards and a mutual insurer model. Some researchers point to the nuclear power industry, which built a liability pool with mandatory insurance, as a possible template for frontier AI. Attorneys advising AI developers or insurers should track policy exclusion language at renewal, review whether existing D&O and cyber coverage responds to autonomous-agent incidents, and watch for the first court rulings applying traditional liability doctrines to AI conduct. The story was reported October 16, 2026, and highlighted that day by MIT Technology Review's newsletter.