Deepfake

A deepfake is synthetic content, image, video or voice, produced by artificial intelligence techniques capable of convincingly imitating the appearance or speech of a real person. What was once the stuff of science fiction has become accessible, with consumer tools able to clone a voice from a few seconds of recording or to animate a face with unsettling realism. Malicious uses are multiplying, namely fraud through impersonation of an executive during a video or phone call, disinformation, market manipulation, and harm to reputation or consent. For insurance, the deepfake upsets several balances, by strengthening the effectiveness of business email compromise, now hard to thwart through the mere recognition of a familiar voice, and by undermining the evidential value of recordings, on which many verification and claims procedures nonetheless rely. It also raises questions of liability that remain largely unresolved, between the author of the forgery, the platform that hosts it and the vendor of the tool that made it possible. The deepfake risk thus illustrates the way in which generative AI creates novel exposures that traditional policies had not anticipated.

An employee receives a video call from what appears to be their chief financial officer, voice and face perfectly recognizable, ordering an urgent transfer. It is in fact a deepfake, and the company only discovers the fraud after the payment has been made.

hypertrucage, média synthétique, deep fake