Trump Administration Declines Safety Testing for Open-Weight AI Models

The Federal government of the United States informed leading artificial intelligence companies on August 4, 2026, that it will not perform safety testing on open-weight AI models. Administration advisers communicated this decision during discussions with the firms.

This move signals a specific regulatory shift toward the development of open-weight models, prioritizing a different oversight approach than that used for closed-source systems. The decision clarifies the administration's stance on the government's role in vetting AI models whose weights are publicly available for modification and deployment.