OpenAI推出GPT-6 Cyber安全研究模型并强调更强对齐
Briefing by News4U · First published
今天，OpenAI在X上表示正在推出GPT-6 Cyber，称这是安全研究的重大进展，将更强能力与更好的模型对齐相结合；该模型支撑了此前公布的Linux内核研究，Mozilla、Palo Alto Networks、Calif、Cisco及OpenAI自身的安全团队已在使用，并向部分Daybreak Red客户开放、后续将扩大访问。The Verge与The Register报道称，OpenAI将其列为准备框架下的Critical级别，ExploitGym得分为48.5%（GPT-5.6 Cyber为32.1%），且OpenAI称在另一项对齐评估中逾99%的情况避免利用范围外服务。 [1] [2] [3]
来源
- @openai (X): Introducing GPT‑6 Cyber: a major advance in security research that pairs stronger capabilities with better model alignment.
@openai (X) · · Excerpt
- OpenAI says its new cybersecurity model is better at not hacking other companies
The Verge · · Full text
- The Register: it avoided exploiting out-of-scope services over 99 percent of the time
The Register · · Excerpt