Trump's AI Testing Framework Excludes Open Models

The Trump administration's framework for assessing potential cybersecurity risks from advanced AI reportedly shows no interest in testing open models. Axios reports that the voluntary guidelines not only exclude open models—meaning anyone can download and inspect their core components—but also explicitly state that the framework cannot be used to restrict open models after they have been released.

The AI testing framework was created following an executive order signed by President Trump in June. This order requested AI companies to share their frontier models with the federal government prior to release to address cybersecurity concerns. The aim was to allow federal agencies to evaluate potential risks associated with these powerful AI systems before widespread deployment.

Critics have raised concerns that the exclusion of open models could undermine the framework's effectiveness. While open models foster innovation and accessibility, they can also be a vector for misuse if not properly scrutinized. Leaving these models outside the official evaluation process may create gaps in security oversight.

The administration's approach appears to rely on voluntary cooperation with developers for proprietary models while abstaining from regulating open-source AI. This strategy raises questions about its sufficiency in managing the broader security implications of rapidly advancing AI technology.