Trump AI Cyber Framework Excludes Open Models
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- The Trump administration's AI cybersecurity testing framework excludes open models entirely and explicitly states it cannot be used to restrict open models once they are released.
- The framework was created after Trump signed a June executive order requiring AI companies to share frontier models with the federal government for pre-release cybersecurity review.
- Anthropic, OpenAI, and Google attended a White House briefing on the finalized framework, which the administration does not plan to release publicly.
- The guidelines include a 30-day government review window and apply only to closed-source models with state-of-the-art capabilities and national security risks.
- The framework fails to define either "state-of-the-art" or "national security risk," despite being designed to assess exactly those concepts.
- Compliance is voluntary, though frontier labs like OpenAI and Anthropic have been actively seeking guidance on how to release models without triggering government restrictions.
Why it matters: Smaller closed-source AI providers trying to stay in the White House's good graces now face a 30-day review with no clear thresholds for what triggers it, while open-source models are off the hook entirely. Frontier labs like OpenAI and Anthropic — who have been lobbying for exactly this kind of guidance — get a structured process they asked for, leaving the voluntary framework's real-world reach narrower than its cybersecurity billing suggests.
