The US government is putting pressure on Meta to submit its AI models for safety and security evaluation. According to Engadget, which cites a New York Times report, Meta is currently the only major AI developer that has not voluntarily handed over its models for government review. Federal authorities want to assess what those models can do and find any potential vulnerabilities.
Every other major player in the AI space has already cooperated. OpenAI and Anthropic are both working with the government to test unreleased models, while Google, xAI, and Microsoft have agreed to give the Center for AI Standards and Innovation early access to their new releases. That agency was created under the Biden administration and is now headed by Commerce Secretary Howard Lutnick. It is staffed with technical experts whose job is to vet AI technology before it reaches the public.
Meta has not flatly refused. A company spokesperson, Francis Brennan, said: “We share the administration’s goal of advancing US leadership on robust and secure frontier AI. While we are working through the details, we hope to sign the agreement soon.” The government has reportedly been sending its requests to the company by email.
The push comes after President Trump signed an executive order on June 2 to create a framework for federal AI evaluation. The government has until the end of July to finalize a formal review process, but the target is clear: companies should give authorities up to 30 days to evaluate new AI systems before they go public. Even without that official process in place, most major companies have been sharing models with the government voluntarily for months. Meta has not.
Meta’s latest model, Muse Spark, launched in April. It has two modes:
- Instant – standard responses for quick queries
- Thinking – a reasoning mode that takes a bit longer to work through a prompt for a more thorough answer
Muse Spark is not considered a frontier model on the level of what OpenAI or Anthropic are putting out, but that hasn’t stopped the government from wanting a closer look. The broader AI industry is under increasing federal scrutiny right now, and no company is too small or too open-source to avoid that attention.
The stakes become clearer when you look at what happened with Anthropic in mid-June. The government ordered Anthropic to block all foreign nationals from accessing its Mythos 5 and Fable 5 models, citing national security concerns. Anthropic responded by cutting everyone’s access to stay compliant. Mythos is a specialized cybersecurity AI available only to Anthropic’s Project Glasswing partners, while Fable 5 was built to bring similar capabilities to a wider audience. Anthropic said Fable 5 outperformed any model it had previously released, which likely explains why the government moved quickly to restrict it.
That episode shows how seriously the current administration is taking AI security, and it explains why Meta’s reluctance to cooperate stands out. The pressure on Meta is not just about one company. It signals that the federal government wants a consistent standard across the entire industry, and companies that don’t get in line voluntarily may eventually be required to do so.