AI CURATED HOT STORY · 11/4/2026
OpenAI recently demonstrated a breakthrough in automated vulnerability discovery using its new cybersecurity agent system, GPT-5.5 Cyber. In a controlled environment, the system scanned approximately 7% of the Linux kernel code (around 2.8 million lines) within just 12 hours, identifying 2,308 candidate vulnerabilities and dynamically reproducing 1,724 of them.
Cross-verified data indicates a high precision rate, with only 3.1% false positives attributed to model hallucinations and a reproduction success rate of 74.7%. Crucially, GPT-5.5 Cyber went beyond detection by automatically generating 948 local privilege escalation (LPE) exploits capable of granting root access to unprivileged users. This end-to-end automation proves that AI can scale complex security tasks from discovery to exploit construction, at an API cost of roughly $316,500.
This development significantly impacts infrastructure security. By adopting a "discover first, prioritize later" strategy and supporting the Patch the Planet initiative, OpenAI aims to accelerate upstream fixes. The event marks a shift where AI acts as an autonomous security researcher rather than just an auxiliary tool, potentially reshaping the landscape of vulnerability mining and defense.
The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.