Trump’s AI testing plan is limited, vague, and excludes open models
The Trump administration's AI security framework entirely bypasses open models and fails to define what constitutes a national security risk.
Stock photo for illustration only, not from the actual event
- The AI framework completely excludes open-source models from testing guidelines.
- It establishes a 30-day grace period for the government to review new models.
- The guidelines fail to define key terms like national security risk.
- Compliance with the framework remains voluntary with no public release planned.
The Trump administration's framework for assessing potential cybersecurity risks posed by advanced artificial intelligence reportedly has no interest in evaluating open models, according to recent reports.
Axios reported that the voluntary guidelines not only outright exclude open models—allowing anyone to download them and inspect their core components—but the framework explicitly dictates that it cannot be used to restrict open models after their public release.
This AI testing framework was established following an executive order signed by President Trump in June, which requested that artificial intelligence companies share their frontier models with the federal government prior to release to address cybersecurity concerns.
Stock photo for illustration only, not from the actual event
Tech companies including Anthropic, OpenAI, and Google reportedly attended a briefing regarding the finalized model testing framework at the White House, though the administration has no current plans to make the framework details public.
The exclusion of open models alongside vague regulatory definitions highlights the ongoing friction between advancing national security and maintaining an open innovation ecosystem. Smaller developers may find navigating these ambiguous voluntary standards particularly challenging.
According to Axios, the framework establishes a 30-day grace period for the government to review new models and will apply exclusively to closed-source AI models possessing state-of-the-art capabilities alongside national security risks.
However, the guidelines fail to define what state-of-the-art or national security risk actually means in this context, presenting a peculiar oversight for an initiative supposedly dedicated to analyzing those exact metrics.
While AI developers are under no legal obligation to comply with the voluntary framework, frontier labs such as OpenAI and Anthropic have actively sought guidance on releasing models without triggering regulatory intervention. The framework's ongoing ambiguity regarding frontier capabilities and national security leaves smaller AI providers in a difficult position as they try to navigate relations with the White House.
Source: The Verge
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment