AI 热点

Google DeepMind 新论文:智能体攻击群仍能攻破防御最强的模型

Rohan Paul更新于 11月13日 23:27

New Google DeepMind paper shows prompt injection is not solved, since a swarm of strong AI attacker agents still tricks even the best-defended models.

AI 导读

Google DeepMind 新论文显示提示词注入尚未解决,由多个 AI 攻击智能体组成的 Agentic Optimizer 仍能骗过防御最强的模型。该系统由 8 组 AI 智能体组成,会测试注入、共享笔记并迭代彼此的最佳尝试,在 100 个工具调用任务上测试,攻击者只求触发一次随机无害的工具调用。最多尝试 320 次时,最强攻击者在最稳健的 Gemini 4 Argon 上攻破 34% 的任务,而开源权重模型 Kimi K3、GLM-5.3 和 DeepSeek-V4.1-Flash 仅抵御住 1% 的任务。

推荐理由

AI 评分 73

论文用多智能体协同攻击测试提示词注入,给出最强攻击者与各模型的具体成功率对比。

主题

标签

正文

New Google DeepMind paper shows prompt injection is not solved, since a swarm of strong AI attacker agents still tricks even the best-defended models.

Prompt injection means hiding instructions in data an AI agent reads, like a tool result, to make it act for the attacker. Human red-teaming attacks now almost never work on recent Gemini models, so the problem can look solved.

They built Agentic Optimizer, 8 teams of AI agents that test injections, share notes, and build on each other's best attempts. They tested it on 100 tool-use tasks where the attacker only wanted one random, harmless tool call.

With up to 320 tries, the strongest attacker beat Gemini 4 Argon, the most robust model, on 34% of tasks. Open-weight Kimi K3, GLM-5.3, and DeepSeek-V4.1-Flash resisted only 1% of tasks.

来源

Rohan Paul