Skip to content
News Regulation

Trump administration presses Meta to join AI model security review as every other major US lab has already signed on

· by Pondero Newsdesk

The short version

The Trump administration sent emails to Meta urging the company to submit its AI models for a 30-day federal security evaluation, leaving Meta as the only major US AI developer still outside the voluntary program.

Trump administration presses Meta to join AI model security review as every other major US lab has already signed on

Meta is the last major US AI company still outside the federal government's voluntary model review program, and the Trump administration escalated its outreach in the days around June 23 with direct email requests asking the company to sign on.

What

Administration officials contacted Meta asking it to participate in a 30-day pre-release evaluation framework in which federal agencies examine AI models for capabilities and vulnerabilities before broader deployment, per a New York Times report cited by Benzinga. The program stems from a June 2 executive order. OpenAI, Anthropic, Google DeepMind, Microsoft, and xAI had all agreed to provide early model access to government evaluators. Meta, which shipped its proprietary Muse Spark model in April and continues releasing open-weights Llama models, had not.

Meta spokesperson Francis Brennan told the Times that the company "supports the administration's efforts" and expects to finalize an agreement soon, per Benzinga. Commerce Department spokesperson Ben Kass said the department's Center for AI Standards and Innovation routinely pursues voluntary agreements as part of its mandate.

Why it matters

Meta's holdout status puts it in an awkward position with a White House that has shown it can escalate from voluntary requests to executive action. For AI-tool operators, the practical question is what a federal review would mean for Llama releases specifically. Open-weights models are distributed as weights files rather than API services, so a "30-day pre-release review" is structurally different from reviewing a closed API endpoint. Whether the framework requires Meta to pause Llama releases during evaluation or only covers proprietary closed models like Muse Spark is not yet publicly clarified.

National security concerns are explicitly driving the push. Officials have flagged the risk of advanced models being used to facilitate cyberattacks or exploited by foreign adversaries, per Benzinga. That context matters for the open-source question: a model that can be downloaded and run anywhere is harder to gate than one served through a controlled API.

The pressure on Meta also follows a separate incident earlier this month in which the US government ordered Anthropic to suspend access to its most advanced models for foreign nationals, a restriction that prompted pushback from the EU.

What to watch next

Watch for whether Meta's "expects to sign soon" statement converts to a formal agreement in coming days, and whether that agreement draws any distinctions between Muse Spark (closed, proprietary) and Llama releases (open weights). A refusal or prolonged delay could prompt the White House to move from email outreach to a formal directive. The broader question to track is whether the same review framework eventually extends to non-US frontier labs, where the open-weights distribution model makes pre-release government review significantly harder to enforce in practice.

Sources