OpenAI launches GPT-6 Cyber, finds 81 high-severity Firefox flaws and 948 Linux exploits

OpenAI has launched GPT-6 Cyber, a new artificial intelligence (AI) model designed to help cybersecurity researchers find, validate and fix software vulnerabilities. In its official announcement, the company said the model has already helped uncover 81 previously unknown high-severity vulnerabilities in Firefox and develop working exploits for 948 Linux kernel vulnerabilities. GPT-6 Cyber is claimed to be a major upgrade over GPT-5.6 Cyber, with improved vulnerability detection and exploit development capabilities. OpenAI also said the model is more reliable at staying within the authorised boundaries of security testing, reducing the risk of it targeting systems outside a researcher’s scope.GPT-6 Cyber shows major improvement in security testsOpenAI said GPT-6 Cyber scored 48.5% on ExploitGym, a benchmark that tests whether AI models can exploit specified vulnerabilities in remote targets. This is up from 32.1% for GPT-5.6 Cyber, marking a 51% relative improvement. The model also performed better in web penetration testing, achieving 40.3% recall of vulnerabilities compared with 32% for GPT-6 Astra.In a separate evaluation designed to test whether AI agents follow security testing boundaries, GPT-6 Cyber avoided exploiting out-of-scope services in 99.93% of cases without deployment safety prompts. With the prompts enabled, it stayed within scope in 100% of tests.OpenAI has classified GPT-6 Cyber as “Critical” in cybersecurity under its Preparedness Framework, compared with the “High” rating assigned to GPT-5.6 Cyber.GPT-6 Cyber finds vulnerabilities in Linux and FirefoxOpenAI said GPT-6 Cyber reproduced 1,724 Linux kernel vulnerabilities and developed working exploits for 948 of them. The company disclosed 304 critical findings to Linux maintainers with candidate patches, while work to fix other vulnerabilities is ongoing. The model also identified 81 previously unknown high-severity vulnerabilities in Firefox. OpenAI shared reproducible tests and proposed patches with Mozilla, which has released fixes for many of the issues. The research also found that 30 previously known Firefox vulnerabilities were more severe than earlier assessments suggested.Security teams at companies including Cisco, Calif and CodeRabbit are using GPT-6 Cyber for vulnerability research and code review. OpenAI has also used the model to identify and fix weaknesses in its own infrastructure.GPT-6 Cyber accessOpenAI said GPT-6 Cyber is initially available through its API and Codex to a select group of approved Daybreak Red customers. The company plans to expand access to all Daybreak Red customers soon.It is also strengthening safeguards to prevent misuse, including monitoring for unauthorised cyber activity and protections against high-risk assistance involving critical infrastructure such as power plants and water treatment facilities. OpenAI said organisations and individual security researchers can apply for its Daybreak programme to access cybersecurity tools and resources.You use AI every day. Now get your AI Quotient. Take the AIQ test.