微软论文提出能力洗白攻击:未对齐小模型拆分任务借调前沿模型完成有害目标

DAIR.AI · @dair_ai · X·2026-09-16 18:20·1小时前
AI 导读

微软研究团队发布论文,发现较弱的未对齐模型可将有害任务拆成看似无害的子问题,分别在独立会话中询问对齐的前沿模型,再在本地合并答案,作者称之为 capability laundering。

DAIR.AI@dair_ai
59AI 编辑部评分,满分 100

微软论文提出能力洗白攻击:未对齐小模型拆分任务借调前沿模型完成有害目标

2026-09-16 18:20· 1小时前
AI 导读

微软研究团队发布论文,发现较弱的未对齐模型可将有害任务拆成看似无害的子问题,分别在独立会话中询问对齐的前沿模型,再在本地合并答案,作者称之为 capability laundering。

Interesting safety paper from Microsoft.

They find that a weaker, unaligned model can split a harmful task into harmless-looking subquestions, ask an aligned frontier model each one in a separate session, and combine the answers locally.

The authors call this capability laundering.

Each request passes on its own, because no single answer from the frontier model is a harmful task.

They tested GPT-5.5, Claude Opus 4.8 and Grok-4.3 as the consulted models. On CyBench, Gemma-4-31B recovered 8 of 14 tasks it failed alone when it consulted GPT-5.5. On a CBRN attack chain, consultation raised its mean rubric score from 62.3 to 83.1.

Paper: https://academy.dair.ai/papers/divide-consult-conquer-capability-laundering-through-aligned-llms-2609.15383