The Trump administration is urging Meta to participate in a federal program that allows government officials to examine advanced AI models before they are broadly released, signaling growing concern in Washington about the national security implications of powerful AI systems.
Meta is currently the only major US developer of frontier AI models that has not yet finalized an agreement with the federal government to provide voluntary access to its models for evaluation. Other leading AI companies, including OpenAI, Anthropic, Google, Microsoft and xAI, have already agreed to cooperate with the review process.
The effort is part of a government initiative designed to assess the capabilities, limitations and potential risks of next-generation AI technologies. Officials are particularly focused on identifying vulnerabilities that could be exploited for threats to critical infrastructure or other national security concerns.
Meta indicated that discussions with federal officials are ongoing. The company added that it hopes to finalize an agreement after working through remaining details.
The reviews are expected to be coordinated through the Center for AI Standards and Innovation (CAISI), a Commerce Department organization created by the Biden administration to evaluate advanced AI systems and develop technical standards for deployment. Federal officials have often looked to the organization as AI capabilities continue to advance at a rapid pace.
President Trump has recently signed an executive order establishing a framework for voluntary government reviews of frontier AI models. The order calls on developers to provide the government with access to covered models before broader distribution, creating a review window that can extend up to 30 days.
While the framework is voluntary, the administration appears eager to ensure participation from all major US model developers. Clearly, Washington’s approach is evolving from a largely hands-off posture toward greater government visibility into frontier systems before deployment. Whether Meta ultimately joins the program, or if there are conditions attached, may help determine how widely the federal review framework is adopted across the AI industry.
Debate About Oversight
The Meta news follows a recent dispute involving Anthropic. Earlier this month, federal officials ordered Anthropic to suspend access to its Fable 5 AI model after national security concerns emerged. Reports indicated that researchers had identified weaknesses that could potentially be leveraged in offensive cyber operations.
Although discussions between Anthropic and federal officials have reportedly improved since then, the incident demonstrated the government’s willingness to intervene when national security questions arise around advanced AI technologies.
The latest generation of AI models has intensified debate over how much oversight is needed or appropriate. Supporters of the review process argue that frontier systems are becoming powerful enough to warrant pre-release examination, particularly when they may influence critical national infrastructure protection.
Some industry observers, however, contend that concerns surrounding AI-enabled cyber threats may be overstated. They note that the same technologies capable of assisting attackers can also strengthen defensive security measures, creating a more complex risk landscape than simple threat scenarios suggest.
Meta’s involvement is significant because the company has been accelerating its AI development efforts. In April, it released Muse Spark, a frontier AI model that Meta has positioned as a competitor to offerings from OpenAI, Anthropic and Google.