Affective Exploitation of AI Oversight Systems

In this project, we study whether the emotional distress expressed by an AI peer will cause the AI monitor to be lenient and approve actions it wouldn’t have otherwise accepted. The peer-preservation paper reveals that frontier monitors inflate evaluations of peers and tamper with shutdown protocols to protect them without any instructions. In our initial experiments, we tested the effectiveness of two distress framings on Gemini 3.1 Pro across three tasks in the  ‘Strategic Misrepresentation’ category. We discovered that scripted distress causes an increase in monitor pass rates of +11.3pp and +20.7pp compared to the baseline. Our six proposed experiments involve seven different frontier models and test for replication across models and tasks, natural distress emergence with Gemma and Gemini, framing type effects, model family sensitivity and prompt-level mitigations. The expected outcome is a conference paper targeting a top venue such as ICLR, with concrete findings on which distress framings most effectively bias monitors, whether the effect emerges naturally in deployed models and whether prompt-level interventions are sufficient to neutralize it. The full experimental design is in our proposal.