Qualification: How Does Adversarial Influence Scale in Multi-Agent Systems?
An honest majority is not a security boundary. In the tested multi-agent groups, initially correct agents defected even when deceivers remained a minority, and susceptibility rose with the deceiver proportion. This evidence qualifies static-majority assumptions and supports governance that measures adversarial influence under the actual composition and protocol. It does not prove the general minority-corruption proposition across voting systems. The evidence is limited to two protocols, groups of at most 21 agents, and descriptive estimates.
Affirmed commentary position. This record extends a source-bound scholarly claim but is not a verbatim paper claim.
Holds when
Current debate
Scholarly basis
Evidence and mapping
Topics
governance-designregulatory-failure
Provenance
Verify