DEV Community

FreyaLi
FreyaLi

Posted on

OpenAI GPT-6 Cyber: The AI Security Model That Built 948 Linux Exploits

infographic

OpenAI has released GPT-6 Cyber, a specialized model rated "Critical" under its safety framework, after an autonomous pipeline helped uncover 81 unknown high-severity Firefox flaws and produced working privilege-escalation exploits for 948 Linux kernel issues at a cost of about 1.2 million dollars in API credits.

Key data points:

  • 948 Linux kernel privilege-escalation exploits developed
  • 81 unknown high-severity Firefox flaws identified and being patched
  • 48.5 percent ExploitGym score, up about 51 percent versus the prior model
  • Only 0.07 percent of tests exceeded authorized scope

Defenders argue that AI-accelerated discovery lets patches ship faster than attackers can weaponize. Critics counter that the same capability lowers the bar for offensive use. The pace at which humans can review and patch critical infrastructure is now the bottleneck.

Read the full story on ThreeRouter AI Daily: https://www.threerouter.com/blog/338

Published via ThreeRouter AI Daily. Cover image generated by AI.

Top comments (0)