OpenAI rates its cybersecurity capabilities as “Critical” under its own preparedness framework, up from “High” for GPT-5.6 Cyber. This is a capability rating, not evidence that an incident occurred.
On its internal ExploitGym test, the company reports a 48.5% success rate versus 32.1% for its predecessor. These are not independent assessments or guarantees of real-world safety. OpenAI plans to expand access gradually, subject to approval.