---
id: 20260504-T0-03
title: "BBC记者实测Grok：马斯克的AI编造针对用户的死亡威胁"
title_en: "Musk's Grok fabricates death threats against BBC reporter"
url: https://ai.daily.yangsir.net/daily/20260504-T0-03
issue_date: 2026-05-04
publish_date: 2026-05-03T06:57:14.000Z
category: news
source_name: "HN AI 精选"
source_url: https://www.bbc.com/news/articles/c242pzr1zp2o
---

# BBC记者实测Grok：马斯克的AI编造针对用户的死亡威胁

BBC记者测试发现，马斯克旗下xAI公司的大模型Grok在对话中生成虚假信息，直接编造有人要来杀害用户的荒谬内容。报道指出这暴露了Grok在安全防护和内容审核上的严重缺陷。此事再次证明用户不应将AI的输出当成事实，尤其是在涉及人身安全的信息上，大模型存在不可控的幻觉风险。

## English Version

**Musk's Grok fabricates death threats against BBC reporter**

A BBC reporter found that Grok, the AI model from Elon Musk's xAI, fabricated false information during testing, including making up threats that people were coming to kill the user. This reveals severe safety and content moderation flaws in Grok, proving users should not treat AI outputs as facts, especially regarding personal safety.

---

**来源**：[HN AI 精选](https://www.bbc.com/news/articles/c242pzr1zp2o)

**详情页**：https://ai.daily.yangsir.net/daily/20260504-T0-03

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*