OpenAI Unveils GPT-6 Cyber Security Model, Finds 81 New Firefox Vulnerabilities

Original title: OpenAI 最强网安 AI 模型:GPT-6 Cyber 登场,性能提高 51%、发现火狐 Firefox 浏览器 81 个新漏洞
88AI Score

OpenAI launched GPT-6 Cyber, a security-focused model that scores 48.5% on the ExploitGym exploit benchmark, a roughly 51% relative gain over its predecessor. In practice it reproduced 1,724 Linux kernel vulnerabilities and uncovered 81 previously unknown high-severity flaws in Firefox, submitting evidence and candidate patches to Mozilla. OpenAI rates the model Critical under its Preparedness Framework and limits initial access to approved Daybreak Red customers.

Full Report

IT之家 November 6 news, OpenAI yesterday (November 5) published a blog post announcing the launch of a cybersecurity-specific model GPT-6 Cyber, which the company claims achieves significant improvements over the previous generation GPT-5.6 Cyber in discovering, validating, and fixing software vulnerabilities, with a relative score improvement of 51% on the ExploitGym benchmark.

OpenAI's most powerful cybersecurity AI model: GPT-6 Cyber debuts, performance improved by 51%, discovers 81 new vulnerabilities in the Firefox browser

The ExploitGym benchmark primarily tests whether a model can exploit specified remote vulnerabilities. GPT-6 Cyber scored 48.5%, a relative improvement of about 51% over GPT-5.6 Cyber's 32.1%. OpenAI also stated that without deployment safety prompts, the model avoided attacking services outside the authorized scope in 99.93% of tests; with prompts enabled, this figure reached 100%.

OpenAI's most powerful cybersecurity AI model: GPT-6 Cyber debuts, performance improved by 51%, discovers 81 new vulnerabilities in the Firefox browser

IT之家 note: The original image only has a black background

In terms of practical results, OpenAI researchers used GPT-6 Cyber to reproduce 1,724 Linux kernel vulnerabilities and developed working privilege escalation exploits for 948 of them; meanwhile, the model discovered 81 previously unknown high-severity vulnerabilities in Firefox, and submitted evidence, reproducible tests, and candidate patches to Mozilla, which has begun releasing fixes.

OpenAI's most powerful cybersecurity AI model: GPT-6 Cyber debuts, performance improved by 51%, discovers 81 new vulnerabilities in the Firefox browser

OpenAI has classified GPT-6 Cyber's cybersecurity risk level under its "Preparedness Framework" as Critical, higher than the previous generation's High. The company emphasized that the model is initially available only to approved Daybreak Red customers, accompanied by safety mechanisms such as abuse monitoring, credential protection, sandboxing, human oversight, and blocking of high-risk behaviors targeting critical infrastructure.

OpenAI stated that frontier models are entering a critical window for "using AI to protect software," and that it will expand the coverage of the Daybreak program in the future to help various organizations, independent researchers, and open-source maintainers conduct vulnerability research, code review, and patch development.