AI is too risky to insure, say people whose job is insuring risk

As the race to implement artificial intelligence accelerates, a troubling question arises: what happens when the technology becomes too risky to insure? Recent reports from the Financial Times reveal that prominent insurers such as AIG, Great American, and WR Berkley are seeking permission from U.S. regulators to exclude liabilities related to AI from their corporate policies. Industry insiders express significant concern regarding the unpredictable nature of AI outputs, with one underwriter referring to them as 'too much of a black box.' The apprehension in the insurance sector is well-founded, highlighted by incidents such as Google’s AI Overview mistakenly accusing a solar company of legal issues, leading to a staggering $110 million lawsuit earlier this year. Similarly, Air Canada faced challenges when its chatbot generated a fictitious discount that the airline had to honor. Moreover, a cyber fraud incident last year highlighted the potential catastrophe of AI misuse, where a digitally cloned version of a senior executive from the London-based design firm Arup was used to siphon off $25 million during a seemingly legitimate video call. Insurers' primary fear is not just large individual claims but the systemic risk posed by numerous simultaneous claims triggered by a single AI error. As one executive from Aon stated, while insurers can manage a significant loss of $400 million from one company, they are ill-equipped to handle a scenario where an AI mishap results in 10,000 losses at once.