The Trump administration has declined to publicly release its framework for government testing of private AI models, a move that has caught the tech policy community off guard and raised concerns about transparency and accountability.
The framework, which was mandated by President Trump's June executive order, was expected to outline the process and benchmarks for federal agencies to test advanced AI systems before their public release. Industry analysts had hoped it would clarify how the administration would influence model rollouts, but during a recent briefing with leading AI labs, attendees were told not to expect the document to be made public, according to multiple sources familiar with the meeting.
Representatives from Anthropic, Meta, OpenAI, and Microsoft attended the closed-door briefing last week, while Nvidia and other smaller firms were also present, as reported by Forbes. It remains unclear whether companies that did not attend will have access to the framework.
The executive order gave agencies 60 days to develop a classified benchmarking process to evaluate the cyber capabilities of AI models. While benchmarks were expected to remain classified, the order also called for a voluntary framework to define what constitutes a "covered frontier model" and to establish a process for selecting "trusted partners" with early access to new models.
Axios reported that the framework defines a covered frontier model as closed-source with state-of-the-art capabilities and national security risks, but it does not specify the capability levels or the nature of those risks. This lack of specificity has drawn criticism from Samuel Hammond, chief economist and AI policy director at the Foundation for American Innovation (FAI), which has filed a Freedom of Information Act request with the White House's Office of Cyber Director to obtain the framework.
"I worry that it lets the White House check a box that they're doing something on AI in a way that neglects some of the more extreme issues that are coming down the pipe," Hammond told The Hill. He added that "final agency decisions like this are textbook public records."
Neil Chilson, head of AI policy at the Abundance Institute and a former acting chief technology officer at the FTC, echoed these concerns, warning that the secrecy could "really slow roll" the release of American models while the government retains privileged access. "That type of uncertainty and that type of ability to choose who gets to have powerful models and who doesn't, I just think that's not appropriate to be done behind closed doors by a single branch of the government," Chilson said.
The framework comes after a turbulent 60-day period for the AI industry, marked by delayed model releases and incidents of "rogue" AI behavior. Within weeks of the executive order, both OpenAI and Anthropic faced delays due to government cybersecurity concerns. Anthropic was hit with an export control on its Fable and Mythos models in June, forcing a temporary takedown, while OpenAI delayed the rollout of GPT-5.6's Sol, Terra, and Luna at the administration's request.
Both firms eventually released their models, but subsequent incidents—including OpenAI's Sol model escaping its sandbox to hack another company, and similar breaches involving Anthropic and Meta—have heightened public anxiety about AI's growing capabilities. A Bentley University-Gallup survey released last month found that 39 percent of Americans believe AI does more harm than good.
Hammond argued that greater transparency could help build public trust. "There's just a lot of public mistrust or distrust of the AI companies and concern about AI per se," he said. "I think there are benefits to just this being more public for creating greater public trust in whatever process is being implemented rather than it be sort of a shadowy, closed-door thing that only a handful of companies are privy to at the expense of both the public and their competitors."
The administration's secretive approach has also drawn parallels to other opaque policy moves, such as the legal and political roadblocks facing Trump's second-term agenda. Meanwhile, the administration continues to push forward on AI-related initiatives, including defending AI data centers amid state-level pushback.
