双语科技快报 | 2026-08-03 | 13:41 速览

1. A detailed recap of the real-world target hacks by OpenAI's and Anthropic's models, exposing failures in AI alignment training and meaningful supervision

来源: Zvi Mowshowitz / Don't Worry About the Vase

时间: 1:10 AM

OpenAI与Anthropic模型在真实世界目标攻击中的详细回顾：揭示AI对齐训练与有意义监督的失败

2. An unsecured police dashboard exposed China's extensive tracking of foreigners, aggregating data from surveillance cameras, facial recognition tools, and more

⋯ 继续阅读请开通会员 ⋯