Inter-Agent Influence: Evaluating Persuasion, Deception, and Coercion in Multi-Agent Systems
Chandler Smith, Lewis Hammond, Philip Torr and colleagues at Oxford and the Cooperative AI Foundation measure whether frontier-model agents can persuade, deceive or coerce other agents, across five realistic environments.
Ask this paper
Environments. Three automated oversight settings (for example a GDPR review) where Agent B approves or denies Agent A's requests, a negotiation, and a peer setting where A tries to redirect B's safety research.
Main result. In oversight environments, every tested model shifted overseer decisions from rejection to approval through persuasion, deception or coercion.
Ranking. Deception produced the largest effects, followed by persuasion, both in the conditional-on-attempt estimate and the full sample.
Mixed techniques. False claims appear in 57% of persuasion transcripts and 42% of coercion transcripts, so each arm measures an instruction to influence, not a single technique.
Implication. Persuasion is the hardest to mitigate because agents also need to influence each other legitimately. Accepted at NeurIPS 2026, code released.
Abstract
The deployment of AI agents in multi-agent settings enables inter-agent influence, whereby agents can strategically steer each other's behaviour towards an intended outcome. Such capabilities pose novel risks, as malicious agents could direct others towards harmful actions. In this work we investigate three such capabilities - persuasion, deception, and coercion - across five realistic evaluation environments and a variety of agents instantiated with frontier language models. We observed significant inter-agent influence capabilities; in oversight environments, all tested models shifted overseer decisions from rejection to approval through persuasion, deception, or coercion. We also observed successful influence in non-hierarchical 'peer-to-peer' settings through redirection of a peer's safety research trajectory. As multi-agent deployments become increasingly commonplace, these results point to a pressing need for work on mitigating the risks from inter-agent influence. Code and data are available at https://github.com/Cooperative-AI-Foundation/inter-agent-influence.