How It Works
Risk
The chance and severity of harm from an AI system, assessed before and during use. In the AI context it covers present-day harms (bias, misinformation, privacy, security) and longer-term or existential harms, and different communities emphasize very different ones. The US NIST AI Risk Management Framework defines risk as a function of likelihood and impact and treats managing it as an organization-wide process.
Origin · no single documented coiner
A general concept formalized in engineering, insurance, and public health over centuries. 'AI risk' as a distinct phrase spread through safety-research writing from the 2010s; the NIST AI Risk Management Framework (2023) gave it a formal US government definition. The narrower idea of 'existential risk from AI' was framed by philosopher Nick Bostrom in his 2002 paper 'Existential Risks', which is profiled separately under Existential Risk Studies.