OpenAI just shipped an AI that completes 95% of offensive-security tasks. Its public model does 1.5%. That gap IS the story. GPT-5.6-Cyber isn't smarter than GPT-5.6 Sol — it's un-sandboxed and handed only to vetted defenders like CrowdStrike, IBM and Cloudflare via a new 'Daybreak Red' tier. The ceiling was never intelligence; it's who gets the refusals turned off. Who decides when the brakes come off? www.contentbuffer.com #AI #Cybersecurity #OpenAI #InfoSec #AIsafety