Using several AI models to assess threats? Log where they disagree, write down...

https://deansinspiringperspective.hexaforgey.com/posts/what-metrics-should-we-track-for-ai-oversight-in-escalations

Using several AI models to assess threats? Log where they disagree, write down the assumptions behind each call, name the person who makes the final decision, and set a follow-up review date that fits the risk