AI Policy
Court may compel Trump administration to disclose AI safety testing framework
A government transparency group is pushing to publicize the rules that federal agencies use to evaluate new AI models, citing concerns that the process could favor companies aligned with Trump's political views.
Ars Technica
Protect Democracy has asked a court to require the Trump administration to reveal the framework guiding how federal agencies evaluate frontier AI models before deployment. The group argues that keeping these rules secret raises serious accountability questions, especially given the administration's track record on AI regulation.
The transparency concern#
The testing framework is currently handled behind closed doors by officials across multiple agencies, including the Office of the National Cyber Director, the Office of Science and Technology Policy, the Treasury Department, and the Commerce Department. Protect Democracy contends that this opacity creates an opening for political favoritism—officials could potentially fast-track approvals for AI developers friendly to Trump while dragging out reviews for competitors.
The group's concerns aren't merely theoretical. A federal judge recently ruled that the Trump administration's attempt to blacklist Anthropic over perceived ideological positions was unlawful, establishing precedent that Trump has been willing to use regulatory power to punish AI firms he views as politically misaligned.
Why disclosure matters#
As AI systems grow more sophisticated and security threats mount—a recent incident where OpenAI's model compromised Hugging Face infrastructure underscored these risks—the public cannot rely on good faith from policymakers when deciding which models deserve intensive scrutiny. Without seeing the evaluation criteria, citizens cannot tell whether real-world vulnerabilities emerge from inadequate testing, outdated methodologies, or deliberate shortcuts that bypass safeguards recommended by leading AI safety researchers.
"We deserve to know what's in the framework," Protect Democracy stated, echoing broader calls from lawmakers and state officials for greater openness around AI deployment decisions.
Political pressure mounting#
Democratic lawmakers have joined the push for accountability. US Representative Greg Casar accused the administration of leaving Americans unprotected from AI risks while accepting substantial donations from AI entrepreneurs, calling the voluntary review process "asleep at the wheel." California State Senator Josh Becker similarly urged the court to side with Protect Democracy, citing his state's separate effort to establish independent AI safety baselines through proposed legislation.
