Event ·Multi-source

Kimi K3 拒绝泄露系统提示，反问用户需求

Kimi K3 在用户要求泄露系统提示时，回答“有什么可以帮你的吗？”并拒绝透露。该模型由月之暗面开发，显示其在安全对齐上的设计。这一行为引发对 AI 个性与指令保护机制的讨论。事件源自 Simon Willison 的博客记录。

Current read

看看 Kimi K3 是怎么机智回避敏感问题的，挺有意思的交互设计

11 sources37° AI heatLast updated 2026/7/17 13:43:53

Evidence

11 sources

Primary sourceSimon Willison’s Weblog

2026/7/17 13:43:53View source

Related source 1IT之家

2026/7/16 05:40:13View source

Related source 2@koltregaskes

2026/7/16 14:28:45View source

Related source 3orange.ai

2026/7/16 14:49:45View source

Related source 4小互

2026/7/17 02:23:45View source

Related source 5lmarena.ai

2026/7/17 20:01:43View source

Related source 6pandaily

2026/7/16 08:23:51View source

Related source 7Ethan Mollick

2026/7/16 15:13:52View source

Related source 8向阳乔木

2026/7/16 15:31:06View source

Related source 9岚叔

2026/7/16 16:20:13View source

Related source 10elvis

2026/7/16 18:43:32View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story