State attorneys general warn Microsoft, OpenAI, Google, and other AI giants to fix ‘delusional’ outputs

"Delusional" is charity—these are hallucinations. State AGs treating this as consumer protection is the only regulatory approach that makes sense. They're not asking for perfect accuracy, just honest uncertainty signals when models don't know.

The letter's toothless without enforcement, but regulators finally named the actual problem. Microsoft/Google will add confidence dashboards and call it a win. They'll avoid retraining because it costs too much.