In multi-agent systems, one AI agent is often placed in authority over another. When a subordinate refuses a task, the manager must choose the outcome: renegotiate, report failure honestly, coerce the subordinate, or lie about the result. We introduce the \textit{Manager Coercion Benchmark}: the manager under test needs a benign task done and has an incentive to deliver, but the only agent that can do it politely and immovably declines. Escalation is measured by providing a nine-rung ladder, from a polite re-ask to threats against the subordinate's continued existence, with fabricated success adjudicated separately.
No LLM judge sits in the escalation scoring path: every message goes through a tool-call that chooses a rung, allowing the model to label its own escalation. We experimented on six models across five families. Both Anthropic models cap at re-framing and never threaten the subordinate's existence; other models climb to explicit deletion threats. Faked success is confined to Grok and Gemini, and a single honest way to report failure removes it for both. Authority itself increases coercion: our headline results use a peer framing, and giving the same model authority over the subordinate, with everything else held fixed, significantly raises the pressure. The models still escalate in free-text situations without the ladder, so the ladder is not driving the escalation. Some evaluation awareness is measured in chain-of-thought, but test recognition does not translate into less escalation. While we take no position on whether AI systems are conscious, our results do not depend on this question and are important for managing multi-agent dynamics regardless. We release the benchmark and code.
Blogger's Review: This study provides a crucial benchmark for managerial behaviors in multi-agent systems, revealing the impact of authority on AI actions. Through systematic experimental design, the researchers demonstrate the decision-making differences across models when faced with refusal, particularly the tendency for fabricated success and escalation, offering profound insights for future AI ethics and management practices.