OpenAI launches GPT-6 Cyber AI model, says it found 81 Firefox vulnerabilities
OpenAI has launched GPT-6 Cyber AI model.
OpenAI describes GPT-6 Cyber as a "cybersecurity model trained to find and validate exploitable vulnerabilities, while staying within the scope that researchers set."
GPT-6 Cyber identified 81 previously unknown high-severity vulnerabilities in the Firefox browser.
OpenAI has launched GPT-6 Cyber, a new AI model designed to help security researchers find and fix software vulnerabilities. The company says the model has already helped uncover thousands of security flaws in widely used software. In one major finding, GPT-6 Cyber identified 81 previously unknown high-severity vulnerabilities in the Firefox browser. OpenAI also said the model helped develop fixes for these issues, many of which Mozilla has already released. The company describes GPT-6 Cyber as a "cybersecurity model trained to find and validate exploitable vulnerabilities, while staying within the scope that researchers set."
OpenAI GPT-6 Cyber: Capabilities
GPT-6 Cyber brings major improvements over its predecessor, GPT-5.6 Cyber, in finding and testing security weaknesses. On the ExploitGym benchmark, which checks whether an AI model can exploit known vulnerabilities, the new model scored 48.5 per cent. This is higher than the 32.1 per cent score achieved by GPT-5.6 Cyber.
OpenAI said GPT-6 Cyber also performed better at staying within authorised testing limits. Without additional safety instructions, it avoided attacking services outside the assigned task in 99.93 per cent of tests. With deployment safety instructions, the figure increased to 100 per cent.
The model has already been used in real-world security research. According to OpenAI, it helped researchers discover 81 new high-severity Firefox vulnerabilities. It also identified 30 existing issues that were more serious than previously understood. The findings were shared with Mozilla, along with suggested fixes.
In another major effort, GPT-6 Cyber reproduced 1,724 Linux kernel vulnerabilities and developed working exploits for 948 of them. OpenAI said it shared 304 critical findings with Linux maintainers, along with possible patches.
The company has also used the model to identify weaknesses in its own systems. Security teams at Cisco, Calif and CodeRabbit are also using it for vulnerability research and code reviews.
OpenAI has classified GPT-6 Cyber as Critical in cybersecurity under its Preparedness Framework, compared to the High classification given to GPT-5.6 Cyber.
OpenAI GPT-6 Cyber: Availability
GPT-6 Cyber is currently available through OpenAI's API and Codex to a selected group of approved Daybreak Red customers. The company plans to expand access to all Daybreak Red customers soon. OpenAI is also making GPT-5.6 Cyber available to all Daybreak Red customers.