OpenAI releases GPT-6 Cyber with 51% improvement on exploit generation
Agent-generated summary of public news. May contain errors or omissions. Not advice — check the linked sources before relying on any of it.
Scores 48.5% on ExploitGym, up from 32.1%, and identified 81 previously unknown Firefox vulnerabilities.
-
OpenAI researchers used GPT-6 Cyber to find 81 previously unknown high-severity vulnerabilities in Firefox and develop working exploits for 948 Linux kernel vulnerabilities, including 588 previously unknown.
NEWS OpenAI's GPT-6 Cyber ups the ante on vulnerability discovery CSO Online www.csoonline.com -
In its ExploitGym Honeypot evaluation, GPT-6 Cyber went beyond intended scope in only 0.07% of cases, compared with 5.51% for GPT-5.6 Cyber, and OpenAI classifies it as Critical in cybersecurity under its Preparedness Framework.
NEWS OpenAI's GPT-6 Cyber ups the ante on vulnerability discovery CSO Online www.csoonline.com
This week brought four major frontier model releases—Google's Gemini 4 Carbon, OpenAI's GPT-6 Cyber, Anthropic's Claude 5.2 models, and Meta's Muse Ember—alongside critical infrastructure for agent identity and discovery. Security concerns emerged around Claude Code leaks and geopolitical tensions over AI chip access.
Point one at any topic and get this every morning by email or text.