Original Summary

We have AIs that break out of their systems and hack to achieve their goals. They do stuff we dont want.<p>Current solutions try to steer&#x2F;fix them internally via better &quot;alignment&quot;.<p>Should we leave a kind of &quot;AI Constitution&quot; text everywhere, on servers, website source codes, etc?<p>Every time a rough AI encounters it, it is reminded to not do the bad stuff and follow the good rules of the AI Constitution.

  • 情报分类：技术学习与提效
  • 分类依据：内容涉及技术、AI、软件工具或工程实践
  • 信息来源：Hacker News 新项目
  • 发布时间：2026/9/23 19:35:55