United States voluntary review framework before completion of release of AI model, OpenAI, Anthropic, etc

On 4 August, the Trump Government completed a front-line AI model review framework and invited OpenAI, Google and Anthropic staff to the White House for discussion. AI may volunteer for government testing before the model is open to partners and the public. The framework is based on an executive order signed by Trump on 2 June. At the time, Anthropic did not have an open front model Mythos for fear of cybersecurity risks. The executive order then requires multiple departments to establish rules within 60 days to determine which models require focused review. The Government would like to work with enterprises to map cyberattacks and national security risks prior to the release of the model, but for the time being no mandatory authorization system has been put in place. The White House states that the framework has been completed and that it has not yet been clarified whether it will enter into force. Previously, the United States Government had restricted the export of the Anthropic model Fable and required OpenAI to publish GPT-5.6 in batches. The new framework will transform these temporary interventions into a fixed set of processes。