Haven't run into a single guard-rail issue with GPT 5.6 Sol like everyone has been complaining about

Guessing there's trust tiers based on peoples usage history

Bigger problem is it's pedantic and hallucinates vulns