OpenAI has introduced GPT‑6 Cyber, a model built to find, validate and help fix exploitable vulnerabilities while staying within researcher-set boundaries. It is initially available through the API and Codex to select approved Daybreak Red customers, with access to all Daybreak Red customers planned soon. Daybreak Red supports advanced exploit development and authorized testing, while Daybreak Blue gives verified defenders fewer unnecessary refusals when validating vulnerabilities and fixes.
Introducing GPT‑6 Cyber: a major advance in security research that pairs stronger capabilities with better model alignment.
— OpenAI (@OpenAI) November 5, 2026
The model powered the Linux kernel research we announced yesterday, and security teams at companies like Mozilla, Palo Alto Networks, Calif, and… pic.twitter.com/34uAeNnhwP
GPT‑6 Cyber scored 48.5% on ExploitGym, up from 32.1% for GPT‑5.6 Cyber, a 51% relative gain. In black-box web penetration testing, it reached 40.3% vulnerability recall versus 32.0% for GPT‑6 Astra. It also completed 95% of RemoteExploitBench tasks against controlled deployments of software including GitLab, PostgreSQL and Jenkins. OpenAI classifies the model as Critical for cybersecurity under its Preparedness Framework, up from High for GPT‑5.6 Cyber.
Alignment training focused on authorized scope, restrictions and accurate reporting. In ExploitGym Honeypot, the model avoided out-of-scope services in 99.93% of tests without a deployment safety prompt, compared with 94.49% for GPT‑5.6 Cyber, and stayed within bounds in 100% with the prompt. OpenAI is pairing the release with cyber-specific monitoring, threat checks, credential protections, IP allowlisting, sandboxing and human oversight.
The model produced reproducible tests for 81 previously unknown Firefox vulnerabilities rated high severity by Mozilla and showed that another 30 known issues were more severe than reported. Mozilla has begun landing fixes. OpenAI researchers also used it to reproduce 1,724 Linux kernel vulnerabilities and develop working exploits for 948, compared with 166 exploited by GPT‑5.6 Sol in the same harness. They disclosed 304 critical findings to maintainers with candidate patches.
OpenAI also used GPT‑6 Cyber in controlled campaigns against its own infrastructure, where agents chained vulnerabilities to escape microVMs and reach underlying hosts, prompting patches and tighter network restrictions. Cisco, Calif and CodeRabbit are among the security teams using it for vulnerability research and code review. The company is also forming a Daybreak Red group for approved critical infrastructure defenders and subsidizing access through its $1 billion Daybreak for Frontline Defenders grant program.
The limited rollout reflects the risk that the same capabilities can support abuse. GPT‑5.6 Cyber is now available to all Daybreak Red customers, while Daybreak Blue alpha access to GPT‑6 Astra is expanding to more organizations and individual researchers.
Sources and related context
- GPT-6 Cyber system card: Supports: OpenAI details the model's Critical cybersecurity classification, authorized-scope evaluations and access safeguards.
- Mozilla security fixes in Firefox 159: Supports: Mozilla documents shipped Firefox security fixes, including high-severity vulnerabilities credited to OpenAI researchers.
