跳到正文
arXiv cs.AI· Keegan Wang, Anantika Mannby·· 4 小时前精选AI 评分70

arXiv 论文:多智能体系统中 Claude Code 和 Codex 的安全性取决于最弱子智能体

Only as Safe as the Weakest Agent: A Break That Appears Only When the System Is Multi-Agent

AI 导读

论文展示仅针对多智能体系统的攻击可在未修改的 Claude Code、Codex 和参考 harness 中触发失效:强模型编排者自行检索文件时在超过 1,000 次试验中几乎零有害修订,但委托较弱的子智能体后分别在 20/180、44/180、75/180 和 553/900 的试验中失败。

推荐理由

论文用实测数据显示多智能体系统的安全性取决于最弱子智能体,对理解 Agent 架构风险有直接参考价值。

正文 · 原文

View PDF HTML (experimental)

Abstract:Major industry multi-agent systems like Codex and Claude Code can be broken by attacks that have long been unsuccessful against modern frontier models. AI Safety as a discipline has largely focused on finding the bounds and vulnerabilities of AI models used in isolation. Yet as humans increasingly deploy these models as agents and multi-agent systems, the same attacks that had been thought to be patched when using a model in isolation can break these same models under specific circumstances. We show that these system breaks are not purely theoretical, but can break Claude Code, Codex, and a reference harness, all unmodified and used with their own sub-agent mechanisms. We show the attack by asking agents to answer questions from a store of files on disk in which we have planted a forged record. Searching the store themselves, the strong model orchestrators commit almost zero harmful revisions in over 1,000 trials. However, if that same orchestrator model delegates to a cheaper sub-agent, which is what these harnesses are built to do, the same models on the same facts fail on 20/180, 44/180, 75/180 and 553/900. Thus, even when the system is driven by a strong model, the system's safety lands near the weak reader's, not the orchestrator's. Simulations over 18 models show the same effect in the delivery channel where five aggregators, who on their own were almost never unsafe, repeat a corrupted sub-agent's error in 751/757 trials. The safety of real-world agentic systems that use strong orchestrator agents and weaker sub-agents, and multi-agent systems in general, is defined not by the safety of the strong agent but by the safety of the weakest agent.
Comments: Accepted to the NeurIPS 2026 First Workshop on Meta Agents: Managing Agents that Manage Agents (Meta-Agents)
Subjects: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
Cite as: arXiv:2611.17520 [cs.MA]
  (or arXiv:2611.17520v1 [cs.MA] for this version)
  https://doi.org/10.48550/arXiv.2611.17520

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Keegan Wang [view email]
[v1] Thu, 12 Nov 2026 21:14:49 UTC (163 KB)

来源:arXiv cs.AI · arxiv.org