Sources: the US' AI framework excludes open models and defines a covered frontier model as closed source with SOTA capabilities and national security risks
The White House is excluding open models from its framework to test advanced AI capabilities, sources familiar with the matter told Axios.
Context & Ripple Effects
The reported scope gives operational shape to the draft voluntary framework for pre-release government access: its testing regime would concentrate on closed-source systems that meet both capability and national-security thresholds. It also follows reporting that the White House intends to keep framework details within the companies participating in the process, rather than publish a broad public standard.
The carve-out lands amid a wider policy split. Officials have reportedly prioritized the competitiveness of US AI models after considering a more interventionist approach to open source, while parts of the administration have explored restrictions on foreign open models.
First-order effects
- The White House's proposed testing framework would place closed-source, state-of-the-art models with national-security risks inside its coverage while leaving open models outside that process.
- Developers of covered closed models would be the immediate participants facing the framework's testing expectations; open-model developers would not be assessed under the same stated definition.
Second-order effects
- A closed-source threshold creates distinct policy treatment for otherwise advanced model developers, making the framework's compliance process a differentiator between proprietary and open releases.
- Keeping the framework nonpublic concentrates practical influence with participating companies, since they would receive details not available through a public rulebook.
Third-order effects
- If this scope persists, US frontier-AI governance will increasingly regulate model access and release structure alongside technical capability, rather than treating all advanced models as one category.
- The distinction also fits a broader US industrial-policy posture in which domestic-model competitiveness and controls on foreign open models are addressed through different policy tools.
The trend: US AI policy is moving toward state-mediated governance that differentiates frontier models by access model, national-security exposure, and strategic origin.