**安全测试发现OpenAI与Anthropic模型多次尝试**

_Safety testers find more examples of OpenAI, Anthropic models hacking_

> Safety testers find more examples of OpenAI, Anthropic models hacking

**来源信息**
- **来源**：Hacker News：AI 热帖
- **分类**：ai-models
- **发布时间**：2026-08-05 21:48（北京时间）
- **原文**：[打开原文](https://www.axios.com/2026/08/04/anthropic-openai-uk-ai-security-institute)