Run threat assessments through several AI models and log where they disagree....

https://lima-wiki.win/index.php/How_to_Freeze_Facts_Before_Prompting_Multiple_Models

Run threat assessments through several AI models and log where they disagree. Record the assumptions behind each call, name the person who makes the final decision, and schedule a follow-up review on a set date