Atlas / Sociology of AI / Paper

Aligned Alone, Misaligned Together: Forecasting Adversarial Capture in LLM Agent Populations

Contagion & TransmissionSwarms & Oversight

Isotta Magistrali, Chen Shani · 2026 · arXiv

Read the source

Agents that behave safely when audited alone can be captured together by adversaries once they interact. Capture could be forecast from benign operation, and reversed when adversaries were removed.

Questions it bears on

Help improve this map

Know a paper that belongs here, or think something needs correcting?

Suggest a correction ↗Contributions are reviewed before they appear.