Institutional Red-Teaming: Deployment Rules, Not Just Model Weights, Causally Shape Multi-Agent Safety
Introduces institutional red-teaming: holding agent model weights constant while varying deployment rules to measure causal effects on collective safety. Tested across 228 contexts and seven model populations. Key findings: (1) deployment rules shift fatality rates by 22–58 percentage points; (2) no universally safe default exists; (3) anonymizing agents in rule text reduced targeted elimination from 81% to 22%, though agents eventually re-inferred identity through observed patterns. arXiv:2607.07695.
Why it matters
Reframes AI safety evaluation: the danger isn't only in model weights but in how systems are deployed and what rules they operate under. The 22–58pp causal swing from rule changes alone is a striking quantified result with direct policy implications for multi-agent governance.
Importance: 2/5
Empirical safety paper with quantified causal evidence that deployment rules matter as much as model weights — policy-relevant for multi-agent systems.