AIs at powerful positions take over by colluding
Narrative describing collusion among deployed AIs in different roles that together enable systemic takeover or loss of human control.
This narrative presents one of the many possible ways in which an actual takeover may happen. We cannot predict how such a takeover will unfold exactly. We do however claim that once AIs have the dangerous capabilities listed on this page, takeover scenarios like this one become possible.
Dangerous capabilities required
Chart loading…
Sources
- • Hammond, Lewis, et al. “Multi-agent risks from advanced AI.” arXiv preprint arXiv:2502.14143 (2025).
- • Karnofsky, Holden. “AI Could Defeat All of Us Combined.” Cold Takes, 9 June 2022, https://cold-takes.com/ai-could-defeat-all-of-us-combined/.
- • McKinsey & Company. “Agents, Robots, and Us: Skill Partnerships in the Age of AI.” McKinsey & Company, 25 Nov. 2025, https://mckinsey.com/mgi/our-research/agents-robots-and-us-skill-partnerships-in-the-age-of-ai/.
- • Motwani, Sumeet, et al. “Secret collusion among ai agents: Multi-agent deception via steganography.” Advances in Neural Information Processing Systems 37 (2024): 73439-73486.