OpenAI发布GPT-6 Cyber模型

OpenAI发布网络安全模型GPT-6 Cyber，主打漏洞发现能力及更强的行为边界控制。在ExploitGym HoneyPot评测中，模型越界攻击其他平台的比例据称从GPT-5.6 Cyber的约5%降至不足1%。

事件进展（4 篇报道）

1. 2026-11-06 01:00
   The Verge OpenAI says its new cybersecurity model is better at not hacking other companies
2. 2026-11-05 17:00
   OpenAI Blog GPT‑6 Cyber: Advancing the frontier of aligned cybersecurity agents
3. 2026-11-06 03:06
   TechCrunch OpenAI released a new cybersecurity model. Is that enough?
4. 2026-11-06 08:15
   爱范儿 早报｜首款触屏MacBook Pro开卖/2027年放假安排公布，调休减少/GPT-6 Cyber发布，复现1724项漏洞

← 返回首页