CRS tried to use AI, but less than 3% of results met their standards, director says - FedScoop

A federal agency found that AI tools produced unusable results over 97% of the time, raising real questions about AI reliability in government work. Important for understanding AI limitations in practice.