The United States begins controlled release of domestic frontier AI

One hundred American institutions, including major corporations and federal agencies, received authorization Friday to access Anthropic’s Claude Mythos 5 model. The decision follows a two-week block imposed by the Trump administration under new export controls, which had effectively suspended the model’s operations. The relaxation of these restrictions signals the emergence of a specific regulatory framework where the U.S. government maintains direct oversight over the distribution and export of frontier-tier artificial intelligence.

The Commerce Department cited significant progress in negotiations regarding safety protocols and jailbreak prevention as the catalyst for the release. Under the new arrangement, specific licenses will no longer be required for the transfer of the model to a list of pre-approved entities, including their foreign national employees. However, the fate of Fable 5, a companion model previously available to consumers, remains undecided. The administration’s posture indicates that access to high-compute models will be treated as a strategic asset, with the government acting as a clearinghouse for domestic and international deployment.

This shift toward active state supervision arrives as OpenAI begins a similar limited release of its GPT-5.6 model to government-approved partners. The simultaneous moves suggest a coordinated effort by Washington to balance national security concerns with the speed required by the global technical race. While U.S. officials maintain that the framework preserves American leadership, the new regime has introduced friction with international allies. European officials have expressed concern over their increasing dependence on Washington’s permit process for access to the tools defining the next decade of industrial and economic competition.