automated-security

STAR-Teaming introduces a multiplex-network-driven multi-agent framework that automates LLM red-teaming, achieving higher attack success rates with lower compute by organizing attack strategies into interpretable semantic communities.