OpenAI GPT-6 Cyber: 7 Things to Know About Its New Cybersecurity AI
Poulami Saha
What GPT-6 Cyber is: GPT-6 Cyber is OpenAI’s cyber-specialized AI model for authorized security work, including vulnerability discovery, exploit development, and web penetration testing.
Major performance gains: OpenAI reports a 51% relative improvement over GPT-5.6 Cyber on ExploitGym, scoring 48.5% versus the earlier model’s 32.1%.
Real-world research results: OpenAI says the model found 81 previously unknown high-severity Firefox vulnerabilities and helped create working exploits for 948 Linux kernel vulnerabilities.
Offense and defense: It can identify flaws, build proof-of-concept exploits, validate exploit chains, and suggest fixes, serving both offensive testing and defensive security teams.
Critical risk classification: OpenAI rates GPT-6 Cyber “Critical” under its Preparedness Framework, its highest cyber-risk tier, because it can develop functional zero-day exploits.
Tighter scope control: The model stayed within authorized boundaries in 99.93% of honeypot tests without special safety instructions, up from 94.49% for its predecessor.
Restricted initial access: GPT-6 Cyber is initially available through API and Codex only to selected approved Daybreak Red customers, with broader Red access planned.