The federal government has reached agreements with Google, Microsoft Corp., and Elon Musk’s xAI to vet unreleased artificial intelligence (AI) models before they reach the public, the National Institute of Standards and Technology (NIST) announced Tuesday.
The deals signal a significant shift in the oversight of frontier AI as officials scramble to contain escalating national security and cybersecurity threats.
The partnership, managed through the Center for AI Standards and Innovation (CAISI) within the Department of Commerce, allows government scientists to evaluate the capabilities and safety risks of cutting-edge systems ahead of their commercial launch. The move comes as the emergence of ultra-powerful models, such as Anthropic’s Mythos, has pushed concerns regarding AI-enabled hacking to a critical tipping point.
“Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications,” CAISI Director Chris Fall said in a statement. “These expanded industry collaborations help us scale our work in the public interest at a critical moment.”
The urgency in Washington has been fueled by Mythos, a model Anthropic describes as being “far ahead” of its peers in cybersecurity capabilities. The tool has caused a wave of anxiety across the banking and utility sectors, leading Anthropic to restrict its release to a small group of vetted organizations while briefing senior U.S. officials on its potential to supercharge cyberattacks. (The Trump administration is considering an executive order that would establish government oversight for advanced AI models.)
Under the new agreements, developers will often provide CAISI with “raw” versions of their models – systems with safety guardrails stripped back – to allow for deep-tier probing of risks related to biosecurity, chemical weaponry, and digital infrastructure. While tech giants like Microsoft regularly conduct internal testing, executives acknowledged that the government provides a unique layer of expertise.
“Testing for national security and large-scale public safety risks necessarily must be a collaborative endeavor with governments,” said Natasha Crampton, Microsoft’s chief responsible AI officer.
The initiative expands upon a framework established under the Biden administration and fulfills a July 2025 pledge by the Trump administration to partner with the private sector on AI vetting. However, the move may also signal a hardening of the current administration’s “light-touch” regulatory stance. Reports from The New York Times suggest the White House is weighing a formal, mandatory review process for new models, though a spokesperson characterized such discussions as “speculation.”
The collaboration also addresses a significant resource gap. Jessica Ji, a research analyst at Georgetown’s Center for Security and Emerging Technology, noted that government agencies often lack the massive computing power and technical manpower that Big Tech possesses. These deals bridge that gap, providing CAISI with the resources necessary to conduct more than the 40 evaluations it has already completed.
As the Pentagon simultaneously moves to deploy advanced AI across classified networks, the tension between rapid innovation and national defense remains high.
With Google and xAI remaining largely silent on the specific terms of the deal, the tech industry now finds itself at a crossroads: balancing the race for AI supremacy against the growing requirement for federal oversight.
