A few days ago, I published High Confidence Is Not a Security Control, having spent some time looking at JEV, a specialised decision model, and the security implications of allowing probabilistic AI judgments to drive consequential actions. The concern was straightforward: a model receives evidence, …