Meta is the only major US AI developer that has yet to submit its models for government safety review, and the Trump administration is pushing hard to change that.
Administration officials have been pressing Meta through email communications to join a voluntary federal review program for advanced AI systems, the New York Times reported Tuesday, citing four people familiar with the matter. The program would let federal agencies evaluate cutting-edge AI models for capabilities and vulnerabilities before broader deployment.
The pressure is part of a wider push by the administration to increase oversight of rapidly advancing AI technologies. President Trump signed an executive order on June 2 directing the government to establish a formal review framework by the end of July, with the goal of giving authorities up to 30 days to evaluate new models before public release.
OpenAI and Anthropic are already collaborating with the government to test unreleased models, Reuters reported. Google, xAI, and Microsoft have agreed to provide the Center for AI Standards and Innovation, an agency created by the Biden administration, staffed with technical experts, and now headed by Commerce Secretary Howard Lutnick, with early access to their new models.
Meta has so far declined to follow suit. The company launched its latest model, Muse Spark, in April.
It features both "Instant" and "Thinking" modes, with the latter enabling reasoning capabilities that take additional processing time for more thorough answers.
"We share the [Trump] administration's goal of advancing US leadership on strong and secure frontier AI," Meta spokesperson Francis Brennan told the Times. "While we are working through the details, we hope to sign the agreement soon."
Commerce Department spokesperson Ben Kass said the Center for AI Standards and Innovation routinely works with companies on voluntary agreements as part of its mandate. The administration's urgency reflects growing national security concerns about whether advanced AI models could enable cyberattacks, support military operations, or be exploited by foreign adversaries. Earlier this month, the government ordered Anthropic to suspend access to its Mythos 5 and Fable 5 models for all foreign nationals, citing national security.
Anthropic complied by blocking everyone's access to ensure compliance.
Meta's Muse Spark isn't as powerful as frontier models from competitors like Anthropic. But the company's reluctance to participate voluntarily puts it in an increasingly isolated position as the rest of the industry cooperates with the government's AI review push.
