Trump Faces Order to Reveal AI Safety Testing Protocols

Breaking: The Full Story

A federal district judge in Washington, D.C., has issued a preliminary ruling that may force former President Donald Trump to reveal internal guidelines used by U.S. intelligence and defense agencies to assess artificial intelligence systems for national security risks. The decision, handed down late Friday by Judge Tanya Chutkan, stems from a Freedom of Information Act (FOIA) lawsuit filed by the Electronic Frontier Foundation (EFF) against the Trump administration. At issue are so-called “black-box” evaluation protocols—classified frameworks that agencies like the Department of Defense and the National Security Agency reportedly use to test AI models for vulnerabilities, bias, and potential misuse before deployment in critical infrastructure or military applications. According to court documents, these protocols have never been publicly disclosed, despite growing congressional and public pressure to standardize AI safety testing across industries.

The lawsuit gained momentum after a 2023 Government Accountability Office (GAO) report revealed that federal agencies were using inconsistent, non-transparent methods to evaluate AI systems, particularly in semiconductor supply chain contexts. A GAO spokesperson confirmed that the report flagged a lack of unified standards, noting that discrepancies in testing could lead to systemic failures in industries reliant on advanced AI—such as autonomous vehicles, semiconductor manufacturing, and defense systems. Among the agencies named in the FOIA request are the NSA, DARPA, and the newly formed AI Safety Institute under the Department of Commerce. Legal experts suggest that if the judge’s ruling stands, it could set a precedent for future transparency in AI governance, especially as the U.S. races to catch up with global peers like the EU, which finalized its AI Act in May 2024.

The timing of this development is particularly sensitive, coming just weeks after the Trump administration quietly expanded its AI safety initiative through an executive order in March 2025. That order directed the Department of Commerce to develop “voluntary” AI safety guidelines for U.S. semiconductor firms, including NVIDIA, AMD, and Intel—companies already under scrutiny for their role in powering AI models that may pose national security risks. During a closed-door briefing with tech lobbyists in April 2025, Commerce Secretary Wilbur Ross was quoted in a leaked memo as saying, “We cannot have our chipmakers flying blind when it comes to AI safety.” The memo, obtained by OpenPress Semiconductor Intelligence, also highlighted concerns about foreign adversaries exploiting gaps in U.S. AI testing regimes.

Industry Impact and Significance

If the federal protocols are disclosed, semiconductor manufacturers would face immediate pressure to align their internal AI safety frameworks with newly public standards—or risk losing government contracts or export licenses. For instance, NVIDIA’s recent dominance in AI accelerators, particularly its H100 and upcoming B100 GPUs, has made it a focal point for federal scrutiny. The company has already begun internal audits of its AI models for bias and security flaws, but sources within the semiconductor industry suggest these efforts are disjointed without clear federal guidance. Financial analysts at Banking With Billy AI, a fintech firm specializing in real-time semiconductor stock analytics, have observed a 7% dip in NVIDIA’s share price over the past two weeks—partly attributed to investor unease over regulatory uncertainty. “The market is pricing in the risk that NVIDIA might have to retrofit its entire AI stack to comply with unknown federal standards,” said lead analyst Priya Kapoor. “That’s a massive capital expenditure with no clear ROI.”

The ripple effects would extend beyond NVIDIA. AMD’s Instinct MI325X and Intel’s Gaudi 3 accelerators are also in the crosshairs, as are smaller firms like Cerebras and SambaNova, which have staked their futures on AI hardware. A senior executive at one of these companies, speaking on condition of anonymity, admitted that “we’ve been flying under the radar for years because the rules were vague. Now that the radar is turning on, we’re scrambling.” The disclosure could also accelerate consolidation in the AI chip sector, as larger players with deeper compliance budgets absorb smaller rivals unable to meet stringent new requirements.

The Bigger Picture

This legal battle is unfolding against a backdrop of intensifying global competition in AI governance. The EU’s AI Act, which entered into force in August 2024, requires mandatory conformity assessments for high-risk AI systems—including those used in semiconductor design and manufacturing. Meanwhile, China has already implemented its own AI safety protocols through the 2022 “Interim Measures for the Management of Generative AI Services,” which impose strict data localization and content moderation rules on AI developers. U.S. officials have privately acknowledged that the lack of unified standards puts American firms at a disadvantage, particularly as export controls on advanced chips tighten. A 2024 report by the Semiconductor Industry Association warned that “inconsistent AI safety frameworks could fragment the global chip supply chain, leading to higher costs and reduced innovation.”

The Trump administration’s push for secrecy contrasts sharply with recent bipartisan efforts in Congress to pass the bipartisan Future of AI Innovation Act, which would establish a federal AI Safety Board with broad powers to set testing standards. That bill, introduced by Senators Chuck Schumer (D-NY) and Mike Rounds (R-SD), has gained traction in the Senate but faces opposition from the White House, which has framed it as federal overreach. Legal scholars argue that the judge’s ruling could force a reckoning: either the administration releases the protocols, or Congress intervenes to codify transparency into law. Either way, the semiconductor industry is caught in the middle of a high-stakes showdown over who controls the rules of AI safety.

Expert Analysis

Dr. Elena Vasquez, a senior fellow at the Center for AI Safety and former DARPA program manager, warns that the current opacity is unsustainable. “We are at a pivotal moment where the absence of clear, publicly vetted AI safety standards is becoming a national security liability,” she says. “If the federal protocols are as rigorous as insiders claim, their release could actually strengthen U.S. competitiveness by giving industry a clear roadmap. But if they’re outdated or inconsistent, we risk embedding flawed assumptions into the very systems that will power our economy and defense for decades.” Vasquez predicts that within 12 months, either the courts or Congress will mandate the creation of an independent AI Safety Standards Board—modeled after the Federal Aviation Administration’s role in aviation safety. “The semiconductor industry should prepare for a compliance cliff,” she adds. “The question isn’t whether the rules will change—it’s how fast companies can adapt before the market decides they’re too risky to bet on.”