OpenAI is restricting the initial public release of GPT-5.6, making the most capable Sol variant available only to partners approved by the White House while officials carry out a cybersecurity vetting.The limitation follows an executive order that allows a voluntary 30-day federal review and an export-control directive that forced Anthropic to take Mythos and Fable offline for restricted access.Critics including Alex Stamos say the government has not substantiated the specific security claims and warn that repeated vetting could undermine US competitiveness with China in advanced artificial intelligence.OpenAI says the pause is temporary, has previewed models to the
Pentagon and trusted collaborators, and plans a phased rollout with broader international access once negotiations with the Administration conclude.