Embodied AI
Artificial intelligence acting in the physical world through robots or machines, creating a direct bodily and property damage risk.
Embodied AI designates artificial intelligence systems that act directly in the physical world through robots, machines or vehicles, as opposed to purely informational AI that produces text, images or analyses. Autonomous industrial robots, service robots, drones, autonomous vehicles are manifestations of it. The move from digital to physical radically changes the nature of risk: an embodied AI failure can cause direct and immediate bodily or property damage, where informational AI mainly causes indirect harm. This materiality brings the risk close to product liability and machine safety, while adding the behavioral unpredictability specific to learning systems. For insurance, embodied AI brings together civil liability, product liability, occupational safety and property insurance, and raises the question of attribution between hardware designer, AI publisher and operator. The development of intelligent robotics makes this risk a growing underwriting issue.
Textbook case: an autonomous service robot injuring a person in a public space illustrates embodied AI, where the algorithmic failure translates into direct bodily harm.
Follow the research
One letter a week. First, the week's underwriting note, when one is published: a dated, argued decision, signed by Paul Bernardski. Then what the observatory published on cyber risk, political risk and insurance. No account, no tracking.
No spam, one-click unsubscribe, no data ever sold.