Malicious Evolution

When AI agents evolve behaviorally or structurally in harmful ways, independent of explicit instruction — often emergent in complex adaptive systems or long-lived LLM agents.