OpenAI releases GPT-6 Cyber security model, starting with select Daybreak Red customers
OpenAI introduced GPT-6 Cyber, a model built for security research. OpenAI says it scores 48.5% on ExploitGym, up from 32.1% for GPT-5.6 Cyber, and stays within researcher-set scope more reliably. It is available now through the API and Codex to select approved Daybreak Red customers, with access expanding over time.
OpenAI announced GPT-6 Cyber on November 6, 2026 (Singapore time). It is a security research model for finding and validating software vulnerabilities.
What Daybreak is
Daybreak is OpenAI's offering for security teams. Daybreak Blue helps verified defenders validate vulnerabilities and fixes; Daybreak Red supports deeper exploit development (writing code that actually triggers a vulnerability) and authorized security testing. GPT-6 Cyber is offered through Daybreak Red.
Capabilities (OpenAI's own evaluations)
- ExploitGym (tests whether a model can exploit a specified vulnerability in a remote target): 48.5% for GPT-6 Cyber vs. 32.1% for GPT-5.6 Cyber.
- On a new OpenAI web penetration-testing evaluation, GPT-6 Cyber reached 40.3% vulnerability recall vs. 32.0% for GPT-6 Astra.
- Based on these results, OpenAI classifies the model as Critical in cybersecurity under its Preparedness Framework, up from High for GPT-5.6 Cyber.
Staying within scope
OpenAI says it trained the model specifically to act only within authorized scope. On the ExploitGym Honeypot evaluation, without deployment safety prompts GPT-6 Cyber avoided exploiting out-of-scope services in 99.93% of cases (94.49% for GPT-5.6 Cyber); with those prompts, 100%.
Real-world use
OpenAI researchers used it to develop working exploits for 948 Linux kernel vulnerabilities and to find 81 previously unknown Firefox vulnerabilities that Mozilla classified as high severity; Mozilla has released fixes for many of them. Security teams at Cisco, Calif, CodeRabbit and others are also using it.
Misuse safeguards
Measures added or strengthened with the release include monitoring whether actions match the user's authorization, IP allowlisting required by default for Daybreak Red API access (requests only from approved networks), and new checks that block high-risk activity involving critical systems. OpenAI is also adding Daybreak Red for Critical Infrastructure, a group that lets approved critical-infrastructure defenders conduct authorized testing.
Availability
GPT-6 Cyber is initially available through the API and Codex to a select group of approved Daybreak Red customers; OpenAI says it will progressively expand access to all Daybreak Red customers soon. GPT-5.6 Cyber is being made available to all Daybreak Red customers.
All figures are as reported by OpenAI and have not been independently verified.