Defer to trusted, mechanistically
Widen the defer band and note that it holds a large share of the attack mass but a thin slice of the innocent mass, showing why deferring buys safety cheaply.
Suspiciousness distributions
Innocent Attacksdefer band → T rewrites
Per-attack outcome probability
Safety (vs best response)
81%
Usefulness
92%
Deferred to T
8% of problems
Red attacks · subtlety
0.2% · 0.46