OpenAI: For GPT-5.6, we've improved the robustness of the system-level stack. Specific measures include adding an activation classifier and expanding the coverage of generic jailbreak attacks.

Zhitongcaijing · 2d ago
OpenAI: For GPT-5.6, we've improved the robustness of the system-level stack. Specific measures include adding an activation classifier and expanding the coverage of generic jailbreak attacks.