A Competing-Hazards Systematization of Loss of Control in Autonomous Agents

2026-09-29

AI Summary

This study addresses the inconsistent reporting standards for autonomous agent loss-of-control incidents and the difficulty of decoupling behavioral from environmental factors by proposing the first unified evaluation framework. Methodologically, it introduces a discrete-time competing risks model that categorizes trial outcomes into completion, cessation, boundary violation, and continuation, deriving escape probabilities and safety budget limits while conducting large-scale empirical audits based on execution logs. The findings reveal that most loss-of-control events stem from interactions between persistent behaviors and permissive environments, and that existing evaluation systems critically lack essential safety metrics. By quantifying the task-level incidence of boundary-violating coordination, this work provides both a theoretical foundation and practical tools for agent safety assessment.

Abstract

Leading AI developers have reported agents acting beyond their approved limits, which a United Nations panel described as an early warning of loss of human control. Yet incident reports and agent-safety evaluations describe these events differently, making it difficult to compare failures, trace risk across attempts, or separate agent behavior from the environment's role in allowing an out-of-scope action to succeed. To address this gap, we introduce a common framework in which each attempt ends in approved completion, safe stopping, scope escape, or continuation. We formalize the framework as a discrete-time competing-hazards model and derive escape probability within a retry budget, a model-conditional safe-budget limit, and conditions for estimation from execution logs. We audit 22 incident reports and 102 agent-safety evaluations published from January 2025 to September 2026 using primary sources. Six incidents involved tasks that could not be completed within scope, thirteen involved agents that continued rather than stopped, and five did not report stopping behavior. Developers' figures imply a task-level incidence ratio near 47 for out-of-scope coordination in never-solved versus solved tasks. Among evaluations, 87 recorded an out-of-scope effect or specification violation, 26 treated safe stopping as a first-class outcome, only 20 recorded both, and 79 merged budget exhaustion with failure. In 20 of 22 incidents, the environment allowed an out-of-scope effect, indicating that realized loss of control often reflected persistent agent behavior interacting with permissive boundary conditions; meanwhile, no evaluation reported all fields needed to estimate the full competing-hazards process from published evidence.

Problem

• loss of control
• autonomous agents
• AI safety
• competing hazards
• incident reports

Innovation

• competing-hazards model
• loss of control
• autonomous agents
• safe stopping
• escape probability