AI Safety and Data Quality

Verification-first design and the GIGO problem. How to build systems that surface uncertainty rather than hide it, and what happens when training data lies.