US Restricts Access to Advanced AI Models Amid National Security Review

OpenAI and Anthropic have begun limited releases of their latest artificial intelligence models to a select group of U.S. organizations approved by the administration of President Donald Trump. According to reports, roughly 100 companies and federal agencies have been authorized by the United States Department of Commerce to access the new systems as part of an expanded national security review process.

OpenAI said it is restricting access to its flagship GPT-5.6 Sol model at the request of the U.S. government, making it available only to a limited group of trusted partners during an initial preview. Anthropic likewise announced it had limited access to its Fable 5 and Mythos 5 models in compliance with government directives related to export controls and cybersecurity concerns.

The move marks an unprecedented level of U.S. government involvement in the rollout of frontier AI systems. Officials say the temporary restrictions are intended to assess cybersecurity and national security risks before broader deployment, while both companies have indicated they expect wider public availability once the review process is complete.