White House AI testing framework excludes open models and lacks key definitions
The Trump administration's voluntary AI security framework will not apply to open models and leaves key terms undefined, according to Axios. The framework, created after a June executive order, introduces a 30-day review period but only for closed-source systems with state-of-the-art capabilities and national security risks.
According to Axios, the Trump administration’s voluntary framework for evaluating cybersecurity risks from advanced artificial intelligence will not cover open models. The guidelines explicitly exclude open-source systems, which anyone can download and inspect, and state they cannot be used to restrict such models once released.
The framework stems from an executive order President Trump signed in June, requiring AI companies to share their frontier models with the federal government before release to address cybersecurity concerns. Officials from Anthropic, OpenAI, and Google attended a White House briefing on the finalized approach yesterday, but the administration has not made the details public.
The framework establishes a 30-day grace period for government review of new models, yet it applies only to closed-source AI with state-of-the-art capabilities and national security implications. The document, however, does not define what counts as state-of-the-art or as a national security risk, which critics find puzzling for an initiative supposedly designed to analyze such matters.
While participation is voluntary, leading labs such as OpenAI and Anthropic have been actively seeking guidance on releasing models without triggering government restrictions. Smaller AI providers that want to stay in the White House’s good books could face uncertainty because the undefined terms leave much open to interpretation.
