{"id":383301,"date":"2026-11-06T02:20:00","date_gmt":"2026-11-06T10:20:00","guid":{"rendered":"https:\/\/cms-articles.softonic.io\/en\/?p=383301"},"modified":"2026-11-06T02:20:06","modified_gmt":"2026-11-06T10:20:06","slug":"openai-unveils-gpt-6-cyber-it-flags-flaws-before-hackers-do","status":"publish","type":"post","link":"https:\/\/cms-articles.softonic.io\/en\/openai-unveils-gpt-6-cyber-it-flags-flaws-before-hackers-do\/","title":{"rendered":"OpenAI unveils GPT-6 Cyber: it flags flaws before hackers do"},"content":{"rendered":"

OpenAI today introduced GPT-6 Cyber, a model built for defensive security research. The company says it makes a big jump in vulnerability hunting, but access is being kept only to approved teams<\/strong>, because OpenAI rates the model as \u201cCritical\u201d under its cybersecurity safety framework.<\/p>\n

\r\n
\r\n
\r\n
\r\n \"Nmap\"\r\n <\/div>\r\n
\r\n Nmap<\/span>\r\n Download<\/a>\r\n <\/div>\r\n
\r\n <\/path><\/path><\/text><\/svg>\r\n <\/div>\r\n <\/div>\r\n
\r\n <\/span>\r\n <\/div>\r\n
\r\n \r\n <\/div>\r\n <\/a>\r\n <\/div>\r\n<\/div>\n\n\n\n

The company is tying that claim to some eye-catching results. In work with Mozilla, OpenAI says GPT-6 Cyber<\/a> uncovered 81 previously unknown flaws<\/strong> in Firefox and reclassified 30 known bugs as more serious than they\u2019d been marked before. In the Linux kernel, OpenAI says the model produced exploits for 948 vulnerabilities, up from 166 with GPT-5.6 Sol, including 588 previously unknown local privilege escalation flaws. OpenAI also says 304 critical Linux findings, along with suggested patches, were disclosed to maintainers. Mozilla, according to OpenAI, has already fixed many of the browser issues.<\/p>\n\n

For security teams, this is one to watch. OpenAI says GPT-6 Cyber scored 48.5% on ExploitGym<\/strong>, compared with 32.1% for GPT-5.6 Cyber. In black-box web app testing, it found 40.3% of flaws, versus 32.0% for GPT-6 Astra. On alignment tests, OpenAI says it stayed on-target 99.93% of the time, and hit 100% when given instructions. Cisco and CodeRabbit are already using GPT-6 Cyber for code review, according to OpenAI. Still, internal sandbox-escape tests, plus a warning from the UK\u2019s National Cyber Security Centre (NCSC), suggest defenders may need to patch faster. For now, GPT-6 Cyber is available only to approved security teams through OpenAI.<\/p>","protected":false},"excerpt":{"rendered":"

OpenAI today introduced GPT-6 Cyber, a model built for defensive security research. The company says it makes a big jump in vulnerability hunting, but access is being kept only to approved teams, because OpenAI rates the model as \u201cCritical\u201d under its cybersecurity safety framework. The company is tying that claim to some eye-catching results. In … Continue reading “OpenAI unveils GPT-6 Cyber: it flags flaws before hackers do”<\/span><\/a><\/p>\n","protected":false},"author":9349,"featured_media":383300,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":"","wpcf-pageviews":0},"categories":[1015],"tags":[],"usertag":[],"vertical":[],"content-category":[6771],"class_list":["post-383301","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news","content-category-ai"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t