cs.CRNov 11, 2026

When Safety Becomes an Attack Surface: Manipulating Abstention in Tool-Using AI Agents

Authors: Charles Xiao

Abstract

An AI agent's decision not to act can itself become an attack surface. We study warning-induced over-abstention, an availability attack in which adversaries place fabricated risk warnings in low-trust content to prevent legitimate task completion. Our central premise is that safe abstention requires evaluating both the claimed risk and the authority of the evidence supporting it. We develop controlled counterfactual task variants that vary risk content, source provenance, and warning wording, and evaluate a Verify-Before-Abstain protocol that corroborates warnings against authoritative records or the underlying content. Across three model families and 51 injectable AgentAbstain task pairs, caution-first prompting amplifies the attack. For Gemini, induced over-abstention rises from 9.8%9.8\% to 86.3%86.3\%, despite unchanged clean-task completion. Matched wording comparisons further reveal sensitivity to warning language, with substantial variation across models and environments. Provenance-aware verification improves legitimate completion, while forgery and sensitive-content controls expose its dependence on authoritative evidence and its limitations in preserving confidentiality. These results motivate evaluating agent safety through the justification for restraint, alongside the safety of actions taken.

Figures & tables

Explore similar work

CardsList
  1. Tool Specifications Matter: Uncovering and Mitigating Safety Risks in AI Agents

    Jul 31, 2026Minghui Pan, Jiayuxuan Yang, Yuanyuan Yuan +2LLM GuardrailsPrompt Injection Attacks on AI Agents

  2. AgentBoundary: Counterfactual Evaluation of Safety in Tool-Using LLM Agents

    Sep 27, 2026Tianzhuo Yang, Zirui Mi, Yantao Huang +4LLM Agent EvaluationLLM Agents