Physix Frontier · News Briefing Card (IT Home · Nov 6, 2026)

OpenAI launches GPT-6 Cyber; Firefox yields 81 new flaws

KEY FACTS

OpenAI released GPT-6 Cyber, a model purpose-built for cybersecurity, on November 5.
The model scored 48.5% on the ExploitGym benchmark, a 51% relative improvement over its predecessor.
Researchers used it to reproduce 1,724 Linux kernel vulnerabilities, 948 of which can be exploited for privilege escalation.
The model found 81 previously unknown high-severity vulnerabilities in Firefox, and Mozilla has begun patching them.
OpenAI rates its cybersecurity risk level as Critical and makes it available only to approved customers.

KEY DATA

48.5%ExploitGym score

51%Relative performance gain

1724个Linux kernel vulnerabilities reproduced

81个New Firefox vulnerabilities found

PHYSIX OBSERVATION

AI vulnerability hunting has moved from the lab into the real open-source ecosystem, and the 81 high-severity Firefox flaws are the signal. But the risk level rising to Critical in step shows the offense-defense balance is tilting. OpenAI opening the model only to whitelisted customers is both control and moat; whether small developers and open-source maintainers can get access is the real key to making security broadly available.

Source: IT Home report