← BackThe Verge · AIEmma RothNov 6, 2026, 01:00 UTC+8OpenAI says its new cybersecurity model is better at not hacking other companiesOpenAI 推出网络安全模型 GPT-6 Cyber,称其漏洞发现能力更强,测试中的越界行为比例从 GPT-5.6 Cyber 的约 5% 降至不足 1%。Read the original ↗ShareCopy linkShare imageExport MarkdownAlso reported by 3 other sourcesSee every follow-up