AI 'silent cover' flagged

Moonstone warned on April 20 that insurers using artificial intelligence in claims handling can create “silent” exposure when policy wording never says whether AI-driven errors are covered. (moonstone.co.za) In insurance, “silent” cover means a policy is vague on a new risk, so a dispute starts only after a loss. Brown Jacobson said “silent AI” arises when policies do not explicitly state whether AI risks are covered or excluded. (brownejacobson.com) Moonstone’s example was a claims desk where chatbots answer broker questions, fraud screeners clear files, and handlers wait for machine data such as vehicle logs before deciding a case. The article said those systems are already becoming common in South Africa. (moonstone.co.za) The gap is not limited to one product line. Swiss Re said AI failures can trigger business interruption, professional liability, directors and officers, employment practices, product liability, and copyright-related claims. (swissre.com) Claims teams face a basic liability question when an employee follows a machine recommendation that turns out to be wrong. Kennedys said that can leave insurers arguing over whether the loss belongs under professional indemnity, product liability, or no cover at all. (kennedyslaw.com) The insurance market has seen this pattern before with cyber risk. Lloyd’s told underwriters in 2019 that policies must be clear on cyber coverage by either excluding it or providing affirmative cover, after years of disputes over “silent cyber.” (assets.lloyds.com) That cyber precedent is now shaping AI discussions. Brown Jacobson said insurers are being pushed to make AI wording explicit in the same way, either by affirming cover, adding exclusions, or attaching endorsements that define the trigger. (brownejacobson.com) The pressure is rising as AI incidents pile up outside insurance. Stanford Human-Centered Artificial Intelligence said reported AI incidents reached 233 in 2024, up 56.4% from 2023. (hai.stanford.edu) Moonstone pointed to recent examples now feeding underwriters’ concerns, including Tesla’s autopilot liability, Air Canada’s chatbot dispute, and South African deepfake scams tied to fake investment promotions. South Africa’s regulator has separately warned that deepfake videos used public figures including President Cyril Ramaphosa, Patrice Motsepe, Leanne Manas, and Paul Mashatile to market unauthorized schemes. (moonstone.co.za) (techfinancials.co.za) The practical fix is less futuristic than contractual. Moonstone said carriers need explicit AI policies, documented use of AI in claims, and governance that can show regulators and customers how a machine-assisted decision was made. (moonstone.co.za)