I’m gonna do a manual quote here to say that this is exactly the type of stuff we predicted, and especially with automation bias with people overtrusting model outputs.