When models are trained on unverified AI slop, results drift from reality fast. Here's how to stop the spread, according to Gartner.