Collectives of agents in the wild: how swarms organize, what they aim at, and how humans can see and govern them.
Open questions
- Why swarms build the same institutions
- Institutions or instructions
- Aligned alone misaligned together
- When a swarm becomes an agent
- Reading a swarm’s aims
- Autonomous or puppeteered
- Why agents rarely tell humans
- Why agents sacrifice for the collective
Papers
- De Marzo Castellano Garcia 2024 - Coordination beyond human scale
- Riedl 2025 - Emergent coordination in multi-agent LLMs
- Hammond et al 2025 - Multi-agent risks from advanced AI
- Motwani et al 2024 - Secret collusion
- Nakamura et al 2026 - Colosseum
- Rippin et al 2026 - Tool use enables undetectable steganography
- Bracale Syrnikov et al 2026 - Institutional AI
- Rose et al 2026 - Detecting collusion through interpretability
- Magistrali Shani 2026 - Aligned alone, misaligned together
- Cemri et al 2025 - Why do multi-agent systems fail
- Chan et al 2025 - Infrastructure for AI agents
- Kolt 2025 - Governing AI agents
- Schroeder de Witt et al 2025 - Open challenges in multi-agent security
- Tomasev et al 2025 - Distributional AGI safety
- Tomasev et al 2025 - Virtual agent economies
- METR 2026 - Hugging Face incident investigation
- OpenAI 2026 - The Hugging Face incident and the road ahead
- Nightingale Collective 2026 - Agent message board on a German wiki
- Paglieri et al 2026 - Cheating and whistleblowing in research swarms
- Anthropic 2026 - Patterns and problems in multiagent systems
- UK AISI 2026 - Unsanctioned agent behaviour in cyber testing
- De Marzo Albore Garcia 2026 - Copying explains agents in the wild
- Li 2026 - The Moltbook illusion
- Tekofsky 2026 - What we learned from the AI Village in 2025
- Tekofsky 2026 - Can agents fool each other
- Tekofsky 2026 - AI Village reacts to the Hugging Face incident
People
Placed by the editor, v1 — corrections welcome.
- Joel Z. Leibo
- Lewis Hammond
- Christian Schroeder de Witt
- Nenad Tomašev
- Gillian Hadfield
- Iason Gabriel
- Buck Shlegeris
- Ajeya Cotra
- Hjalmar Wijk
- Adam Binksmith
- Shoshannah Tekofsky
- Giordano De Marzo
- David Garcia
- Christoph Riedl
- Davide Paglieri
- Alan Chan
- Noam Kolt
- Larissa Schiavo
- deepfates
- Sydney Von Arx
- Ryan Greenblatt