White House AI framework reportedly excludes open-weight models
The White House’s newly completed voluntary framework for reviewing advanced artificial intelligence models will reportedly exclude open-weight systems, according to reports on private discussions with AI companies Tuesday, August 4.
The reported decision would make the administration’s pre-release safety process focus on closed, proprietary frontier models with advanced capabilities in areas such as cybersecurity and hacking. The White House said Monday, August 3, that it had completed the framework required by President Donald Trump’s June 2 executive order, but it has not released the document publicly.
What the reported framework would cover
Axios reported that administration officials described a covered frontier model as a closed-source system with state-of-the-art capabilities and national-security risks. The report said the framework does not publicly define what qualifies as “state-of-the-art” or how national-security risk would be measured.
During discussions with AI companies on August 4, the administration reportedly said open-weight models would not go through the same voluntary safety tests. Reuters separately reported that administration advisers told developers those models would be excluded from the testing process.
Open-weight models are systems whose core model weights or other components are made available for download or reuse. That differs from closed systems that remain under a company’s control and are typically accessed through the company’s infrastructure or an application programming interface.
The distinction matters because a company can provide temporary access to a closed model before release. Once an open-weight model is distributed, the original developer may have less ability to control who uses it, how it is modified or where it is deployed.
Review would be voluntary and temporary
The reported process would give the federal government access to covered models for up to 30 days before public release. Axios reported that models would be stored in high-security environments and that detailed logs would track who accessed them.
The framework is not a public regulation, and it does not create a general federal licensing system. Executive Order 14409, issued June 2, directs agencies to develop a voluntary framework with AI developers and to establish secure early access for trusted government partners.
The order also calls for a classified benchmarking process to assess advanced cyber capabilities and determine when a model qualifies as a covered frontier model. It expressly says the order does not authorize mandatory federal licensing, pre-clearance or permitting for the development, publication, release or distribution of AI models.
What remains unknown
The White House has not publicly confirmed the reported open-weight exclusion in an official framework document. Because the full framework and its detailed thresholds remain private, the public cannot yet determine exactly which models or developers would be included.
The reporting also does not establish that every open-source or foreign-developed model is categorically exempt. It specifically concerns open-weight models discussed during private meetings.
That leaves an unresolved policy question: whether a voluntary review process aimed at closed models can address comparable cybersecurity risks from powerful systems whose weights are released for wider reuse. Critics may view the distinction as a potential gap in federal oversight, but the scope and consequences cannot be fully assessed until the administration publishes guidance or clarifies the definitions.
What developers and users should watch
Businesses and developers using an open-weight model should not treat exclusion from the reported process as federal safety approval. The framework, as described, would be voluntary and would not certify models that fall outside it.
The next questions are whether the White House releases implementation guidance, identifies the officials responsible for reviews, explains the threshold for coverage and describes how open-weight systems will be handled when they present similar cybersecurity risks.
The administration’s separate national-security policy directs agencies to use and test both commercial and open-source AI technologies. That policy provides context for the debate, but it does not publicly resolve how open-weight models will be treated under the new pre-release framework.
Sources
- Axios: “Scoop: Inside Trump’s AI framework”
- Reuters: “Trump advisers tell AI firms they will not safety-test open-weight models”
- White House: Executive Order 14409
Discover more from Interactive News
Subscribe to get the latest posts sent to your email.
