---
id: 20261008-T0-15
title: "MLLMs Fail to Refuse when Using Tools Agentically"
url: https://ai.daily.yangsir.net/daily/20261008-T0-15
issue_date: 2026-10-08
publish_date: 2026-10-07T04:00:00.000Z
source_name: "arXiv cs.AI"
source_url: https://arxiv.org/abs/2610.03938
---

# MLLMs Fail to Refuse when Using Tools Agentically

arXiv:2610.03938v1 Announce Type: new Abstract: Agentic multimodal large language models (MLLMs) have recently pushed the frontier of visual reasoning by calling tools such as zooming and tagging. Despite the recent strong success of agentic MLLMs, this work uncovers a critical safety failure in the

---

**来源**：[arXiv cs.AI](https://arxiv.org/abs/2610.03938)

**详情页**：https://ai.daily.yangsir.net/daily/20261008-T0-15

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*