OpenAI postponed the GPT-5.6 launch in June at White House request following security reviews of its capabilities in cybersecurity, biology, and autonomous tasks [Deutsche Welle]. The administration cleared the model after technical testing and also ordered then eased restrictions on Anthropic's Fable and Mythos models [Deutsche Welle]. OpenAI plans a limited preview followed by global rollout and stated that government access processes should not become permanent [Bloomberg][Deutsche Welle].