The Trump administration has decided to keep its new framework for government testing of private AI models out of public view [1].
This move creates uncertainty for technology companies and policymakers who rely on transparent benchmarks to ensure safety and compliance before AI tools reach the public. By withholding the criteria, the government maintains sole control over the evaluation process for private-sector innovations.
The framework is intended to establish specific processes and benchmarks for how the U.S. government tests private AI models before they are released publicly [1]. This internal system allows federal agencies to vet the capabilities and risks of new software without disclosing the exact metrics used for approval.
The decision caught the tech policy sector off guard [1]. Many industry observers had been awaiting the details of the framework to align their development cycles with federal expectations. The lack of transparency means companies may not know exactly which safety or performance hurdles they must clear to satisfy government testers.
While the administration has not provided a public justification for the secrecy, the framework remains a closed-door operation [1]. This approach deviates from previous efforts to create collaborative, open-source standards for AI safety across the public and private sectors.
“The Trump administration has decided to keep its new framework for government testing of private AI models out of public view.”
The shift toward a closed-door testing regime suggests a move away from the 'co-regulation' model of AI governance. By keeping benchmarks secret, the administration can prevent companies from 'gaming' the tests to achieve passing scores, but it also removes the public's ability to audit the government's safety standards.
