Insurers launch cover for losses caused by AI chatbot errors

Insurers at Lloyd’s of London Launch AI Malfunction Coverage

Companies can now protect themselves from losses caused by malfunctioning artificial intelligence tools, thanks to a new product launched by insurers at Lloyd’s of London. The move comes as the sector seeks to capitalize on concerns surrounding the potential risks and errors associated with AI technology.

Armilla, a start-up backed by Y Combinator, has developed policies that will cover companies for court claims arising from harm caused by underperforming AI tools. These policies, underwritten by several Lloyd’s insurers, will cover costs such as damages payouts and legal fees.

While AI has been widely adopted by companies to enhance efficiency, incidents of embarrassing and costly mistakes have highlighted the need for insurance coverage. For example, customer service bots have been known to make errors due to flaws in AI language models, leading to what is referred to as “hallucinations” or the fabrication of information.

Notable incidents include Virgin Money’s AI chatbot reprimanding a customer for using the word “virgin” and DPD disabling part of its bot after it swore at customers. In another case, Air Canada’s chatbot made up a discount, leading to a tribunal ordering the airline to honor the reduced price.

Armilla’s insurance policy would cover losses resulting from such errors, providing companies with the confidence to adopt AI technology without the fear of breakdowns. Unlike general technology errors and omissions policies, which often have low limits on payouts for AI-related liabilities, Armilla’s policy focuses specifically on AI-related losses.

AI language models are dynamic and “learn” over time, but typical technology errors and omissions policies may not cover losses stemming from errors caused by this learning process. Armilla’s policy, however, kicks in when the AI performs below initial expectations, as determined by the insurer.

For instance, if a chatbot initially provided correct information in 95% of cases but later dropped to 85%, Armilla’s insurance could compensate for the degradation in performance. Tom Graham, head of partnership at Chaucer, an insurer at Lloyd’s underwriting Armilla’s policies, emphasized the importance of selecting AI systems with a low risk of breakdowns.

Overall, the launch of AI malfunction coverage by insurers at Lloyd’s of London represents a significant step towards addressing the risks associated with AI technology and providing companies with the necessary protection and peace of mind.