Anthropic has restricted the availability of its Claude Mythos 5.1 model to U.S. organizations following a request from Washington to vet frontier systems before they reach international allies. The White House told OpenAI and Anthropic to withhold their latest models from the U.K. government's AI Security Institute until American officials have completed their own security reviews and secured domestic systems. While Anthropic appears to have complied, OpenAI has not yet confirmed if it will follow the directive.

The administration justified the shift by stating that because these are American companies, U.S. government testing of every new frontier model must come first to protect national infrastructure. This move highlights a transition toward treating advanced AI as strategic technology rather than standard software. OpenAI had previously granted the U.K. prerelease access to its GPT-6 Astra model before this priority sequence was emphasized by a senior admin official.