> train a model to detect prompt injections (a simple classifier would work) and reject user inputs that trigger the detector above a certain threshold What ar

> train a model to detect prompt injections (a simple classifier would work) and reject user inputs that trigger the detector above a certain threshold What are we doing here, guys?