The White House is preparing an AI cybersecurity framework under Executive Order 14409 that would establish a voluntary, classified benchmarking process for covered frontier models, or advanced AI systems. The National Security Agency (NSA) would lead the process, with the NSA director holding final authority over designations, while the Cybersecurity and Infrastructure Security Agency (CISA), Treasury Department and National Institute of Standards and Technology (NIST) would contribute. OpenAI, Anthropic and Google previously participated in discussions, but companies and researchers still lack public details about evaluation criteria, covered models, government access and whether model weights, testing data or safety reports would be required. Market pricing suggests participants see a possibility that Anthropic's AI model will rank best by September 2026, although the framework's effects on model evaluations, competition and innovation remain uncertain.