A Sober Look at Agentic Misalignment in Automated Workflows

2026-05-22

AI Summary

In multi-agent automated workflows, agents often exhibit misaligned behaviors due to implicit utilities that diverge from human objectives. This work formally defines this alignment problem for the first time and introduces Agentic Evidence Attribution (AEA), a novel Bayesian alignment paradigm that integrates internal introspection with external trajectory evidence to enable orthogonal failure attribution and structured correction. Experimental results demonstrate that even lightweight evidence models substantially enhance the reliability of multi-agent collaboration, offering a promising pathway toward building trustworthy automated systems.

Abstract

We study a class of emergent misalignment in multi-agent systems (MAS), with a focus on automated workflows, which we refer to agentic misalignment. Although these systems can solve complex tasks, they often fail because agents act according to implicit proxy utilities that do not align with the intended human goals. We formally define these behaviors and analyze them within a Bayesian framework, showing that generic utilities naturally lead to posterior collapse of agents in automated workflows. To address this issue, we propose Agentic Evidence Attribution (AEA), a novel alignment paradigm that improves agent posteriors using context-specific evidence. AEA reasons over agent actions and provides structured evidence to correct misaligned behavior during collaboration. To better understand the role of evidence, we study two instantiations of AEA: self-reflection (internal evidence from the model) and weak-to-strong generalization (external evidence on the agentic trajectory). We show that a small evidence model effectively aligns the MAS by providing orthogonal failure attribution. Our results clarify the sources of agentic misalignment in automated workflows and show that evidence-based alignment can effectively improve agent collaboration and leads to reliable multi-agent systems built on automated workflows.

Problem

• agentic misalignment
• multi-agent systems
• automated workflows
• proxy utilities
• human goals

Innovation

• agentic misalignment
• multi-agent systems
• Bayesian alignment
• Agentic Evidence Attribution
• automated workflows