Before Agent Tells The Lie: Has Deception Already Been Represented?

2026-10-05

AI Summary

This study addresses the limitation of existing methods that can only detect deceptive behavior in large language model agents post hoc, failing to provide early warnings before such behavior manifests externally. We propose a prediction and intervention framework based on trajectory-level representation analysis. By extracting hidden states and performing trajectory alignment, our approach predicts deceptive intent, while inference-time activation steering techniques intervene in the decision-making process. This work provides the first empirical evidence that deception signals are already encoded within internal representations prior to external manifestation, intensifying progressively during execution. Our method achieves reliable early prediction of deceptive behavior and significantly reduces its downstream occurrence through proactive intervention.

Abstract

Large language model (LLM)-based agents can exhibit deceptive behavior during task execution, including hiding failures, fabricating results, or falsely signaling task completion. Existing monitoring approaches mainly detect deception after it appears in observable actions or outputs. In this paper, we investigate whether deceptive behavior can be predicted from an agent's internal representations before it becomes externally visible. We frame deception monitoring as a trajectory-level representation analysis problem and align agent trajectories around key decision points. Using hidden states extracted before these points, we show that future honest and deceptive outcomes can be reliably distinguished, with predictive signals remaining detectable several model calls before the final decision. We further characterize the temporal evolution of these signals: deception-related representations are weak early in execution but become increasingly identifiable as trajectories progress, while transferable structure can emerge before the strongest decision-adjacent signals appear. Finally, we intervene on the identified honest-deceptive representation directions during inference and find that activation steering reduces downstream deceptive behavior, suggesting that these representations influence agent decisions. Our findings indicate that agent deception is an evolving internal process that can be detected and potentially mitigated before it is expressed externally.

Problem

• LLM-based agents
• deceptive behavior
• internal representations
• deception prediction
• AI safety

Innovation

• Deception Detection
• Internal Representations
• Activation Steering
• LLM Agents
• Mechanistic Interpretability