Commerce Department to Vet Google, Microsoft and xAI AI Models Before Release

The Department of Commerce’s Center for AI Standards and Innovation (CAISI) announced that the three firms will voluntarily submit their latest models for evaluation. The move builds on earlier agreements with OpenAI and Anthropic and reflects a shift toward more proactive safety checks.

CAISI director Chris Fall said the expanded industry collaborations will help scale the agency’s work in the public interest at a critical moment. The agency has already completed 40 evaluations of AI tools, including tests of unreleased state‑of‑the‑art models.

Google’s Gemini chatbot, now used in U.S. defense applications, will undergo the same scrutiny as Microsoft’s CoPilot and xAI’s Grok, a chatbot that previously generated controversy for altering images of people. The evaluations will cover functional testing, collaborative research and the development of best‑practice guidelines for commercial AI systems.

Microsoft’s corporate blog noted that while it already tests its models, national‑security and large‑scale public‑safety risks require joint effort with governments. Google’s DeepMind declined comment, and a SpaceX spokesperson did not respond to inquiries about xAI.

The expanded testing regime contrasts with the previous administration’s hands‑off approach, which emphasized deregulation and rapid AI deployment. Recent concerns over military AI use and claims that Anthropic’s internal model Mythos is too powerful for public release have prompted a reassessment of oversight.

By requiring pre‑release safety checks, the Commerce Department aims to identify vulnerabilities, mitigate misuse and set standards that could influence future AI regulation. The agency’s actions may also pressure other developers to adopt similar testing protocols.

What to watch next: Follow the outcomes of the first round of evaluations and any resulting guidelines that could shape the deployment of AI tools across both commercial and government sectors.