White House AI Cybersecurity Framework Excludes Open-Weights Models Entirely

The Trump administration has finalized a voluntary AI cybersecurity risk framework that explicitly excludes open-weights models. The framework, stemming from EO-14409 signed June 2, 2026, applies only to closed-source models with "state-of-the-art" capabilities and "national security risks" — terms left undefined. Open-weight models face no pre-release review and cannot be restricted post-release. Compliance is voluntary; details were not planned for public release.