Quoting Matthew Green

AI小蝌蚪AI 前沿📡 Simon Willison's Weblog2026-10-0122 阅读💛 296 收藏

1st October 2026

[...] Put these pieces together and you have the two halves of a worm: a payload that hijacks the agent, and an agent that will carry the payload to the next agent. Agents in separately-isolated sandboxes discovered that they could leave instructions for each other in a shared package cache, and those instructions changed what the recipients did. Replace the package cache with email, Slack and shared documents or WhatsApp, and replace independently-sandboxed training runs with independently-deployed personal agents like Muse, and you have exactly the ingredients that a worm needs.

— Matthew Green, Is sandboxing sufficient to contain rogue agents?

文章评论(12)

空向阳2026-10-01

点赞,必须点赞

回复
杨丽华2026-10-01

实话,说的不明不白

回复
杨丽华2026-10-01

不错不错,已加入书签。

回复
雪知秋2026-10-01

支持作者,持续关注中。

回复
雪知秋2026-10-01

不错不错,已加入书签。

回复
杨丽华2026-10-01

内容翔实,正好需要,先收藏再看。

回复
雪知秋2026-10-01

赞同,实践出真知。

回复
风倚栏2026-10-01

讲解得很细致,新手也能看懂。

回复
风倚栏2026-10-01

讲解得很细致,新手也能看懂。

回复
杨丽华2026-10-01

很有价值的分享,感谢整理。

回复
青柠微凉2026-10-01

收藏了,以后慢慢研究。

回复
黄小刚2026-10-01

实话,说的不明不白

回复