The Trump administration has completed a voluntary AI testing framework to assess cybersecurity risks, but it excludes open-source models and lacks clear definitions of key terms like "state-of-the-art" and "national security risk." The framework gives the government 30 days to review new closed-source models before release, though the vague criteria could create confusion for AI companies trying to stay compliant.
What happened
The Trump administration finalized a voluntary framework for assessing cybersecurity risks from advanced AI, following an executive order signed in June. The framework sets a 30-day grace period for government review of new closed-source models but explicitly excludes open models (software anyone can download and inspect) and says it cannot be used to restrict them after release.
Why it matters
The framework does not define what "state-of-the-art" or "national security risk" means, creating ambiguity for AI companies trying to comply. Frontier labs like OpenAI and Anthropic have been seeking guidance on how to release models without triggering government restrictions, but the vague terms could make it unclear which models qualify—a particular problem for smaller providers.
What to watch
The Trump administration is not planning to release the framework details publicly, though AI companies including Anthropic, OpenAI, and Google attended a White House briefing about it yesterday. Compliance is voluntary, so whether labs will meaningfully adopt the framework remains uncertain.
President Trump signed an executive order in June requesting that AI companies share frontier models with the federal government before release, citing cybersecurity concerns. Following that order, the Trump administration finalized a voluntary testing framework to assess potential cybersecurity risks posed by advanced AI. Axios reports that AI companies including Anthropic, OpenAI, and Google attended a White House briefing about the finalized framework yesterday, though the administration is not planning to release the framework details to the public.
The framework sets a 30-day grace period for government review and applies only to closed-source AI models with state-of-the-art capabilities that pose national security risks. However, it does not define what "state-of-the-art" or "national security risk" actually means in context—a significant gap for an initiative explicitly designed to evaluate such risks. The framework also explicitly excludes open models (software that anyone can download and inspect) and states it cannot be used to restrict open models after they have been released.
Because compliance is voluntary, there is no legal obligation for AI companies to participate. Yet frontier labs like OpenAI and Anthropic have been actively seeking government guidance on how to release models without triggering restrictions. The absence of clear definitions compounds the challenge: smaller AI providers wanting to stay in the White House's favor face ambiguity about which of their models would fall under the framework's scope, making it difficult to plan releases confidently.
The framework emerged from an executive order President Trump signed in June, which requested that AI companies share frontier models with the federal government before release to address cybersecurity concerns. However, the finalized version raises questions about its practical utility. By excluding open models entirely—and explicitly forbidding the framework to be used to restrict them after release—the administration has carved out a large segment of the AI ecosystem from oversight. Open models have grown in prominence and accessibility, so their exclusion limits the scope of the cybersecurity review to only proprietary systems.
The lack of clear definitions for "state-of-the-art" and "national security risk" introduces substantial ambiguity. A framework meant to guide companies toward compliance cannot do so effectively if the core terms remain undefined. This is especially problematic for smaller AI providers hoping to stay on the right side of the White House; without clear thresholds, they cannot confidently determine whether a model they plan to release will be subject to the 30-day review period. Frontier labs like OpenAI and Anthropic have been seeking government guidance partly because they want to avoid triggering restrictions, but the vague criteria make it harder for even large companies to navigate the rules.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
The AI news that matters, in one minute each morning.
Sign up free