OpenAI is reportedly preparing a limited GPT-5.6 preview after a White House request tied to AI security concerns. The episode turns frontier model launches into a live test of voluntary federal review, trusted-partner access, and the business cost of slowing powerful AI systems before broad release.