8h ago
OpenAI warns of high cybersecurity risk as performance jumps 27%–76% in Aug–Nov 2025
OpenAI said on December 11, 2025 that its next‑generation models may reach high cybersecurity capability, enabling functional zero‑day exploits and complex attacks. The company cited a capture‑the‑flag performance jump from 27% (GPT‑5, August 2025) to 76% (GPT‑5.1‑Codex‑Max, November 2025). It introduced Aardvark, trusted defender access, and a Frontier Risk Council.