Threat assessment tools built on AI can look confident while being wrong....

https://city-wiki.win/index.php/What_a_Good_Checklist_for_an_AI-Assisted_Decision_Record_Looks_Like

Threat assessment tools built on AI can look confident while being wrong. Here's how we handle it: when two or more models disagree on a risk rating, we log the gap instead of picking the louder answer