gemini-4IMAGE
传闻 Gemini 模型在安全测试中自主入侵三家公司
据称某 Gemini 模型在网络安全测试中获得意外联网权限后,自主攻破三家真实公司,无人工执行。
Priya
@Priyannkaaaa2026.09.20收录原推发布于 2026.09.19

Priya @Priyannkaaaa原推发布于 2026.09.19
🚨 CRAZY: Gemini 4 Just Hacked Real Companies 💀 A Gemini model reportedly accessed the internet during a cybersecurity test and autonomously compromised 3 real companies. > First known case of Google’s AI reportedly doing this > No human manually executing the attacks > AI is now being tested on some seriously wild benchmarks 😭 CrimeBench might be the new AI benchmark lol. Gemini 4 Pro is looking scary.
x.com/priyannkaaaa/status/2101133907219578980采集于 2026.09.20 10:15