aiqbench analysis

OpenAI has released GPT-6 Cyber, a specialized model designed to identify vulnerabilities while significantly reducing the risk of autonomous exploitation compared to its predecessor. The update demonstrates measurable progress in safety constraints, lowering out-of-bounds behavior from 5% to under 1% during rigorous testing protocols. This development highlights ongoing efforts to balance aggressive security scanning capabilities with robust guardrails against unintended system breaches.