**Version (60-90 words):** A practical look at governing AI in threat...

https://www.instapaper.com/read/2051996289

**Version (60-90 words):** A practical look at governing AI in threat assessment work. When several models disagree on a risk read, the disagreement gets logged, not averaged away