U.S. Expands Pre-Release Testing of Google, Microsoft and xAI AI Models
Commerce's CAISI will conduct pre-deployment evaluations of Google DeepMind, Microsoft and xAI models to probe national security and public safety risks and build on prior 2024 partnerships with OpenAI and Anthropic.
BBC NewsUS to safety test new AI models from Google, Microsoft, xAI
Epoch TimesGoogle, Microsoft, xAI Will Allow Government to Vet New AI Models for Security Risks
Fox BusinessTrump admin to review AI models from Google, Microsoft, xAI ahead of public release
CNBCTrump admin moves further into AI oversight, will test Google, Microsoft and xAI models
Overview
The Commerce Department’s Center for AI Standards and Innovation announced on Tuesday that Google DeepMind, Microsoft and xAI will allow CAISI to conduct pre-deployment evaluations of their new AI models before public release.
The agreements expand CAISI’s role in testing frontier AI and build on 2024 partnerships with OpenAI and Anthropic, according to CAISI statements.
Microsoft said it already tests its models and called government collaboration essential, while CAISI Director Chris Fall said the expanded collaborations help scale public-interest work.
CAISI said it has completed more than 40 evaluations, and the Pentagon recently agreed with seven tech firms to use AI across classified networks, underscoring national security stakes.
The White House is weighing creation of an AI working group and potential executive order to vet models, and CAISI said it will pursue collaborative research and classified-environment testing with interagency feedback.
Analysis
Center-leaning sources frame the story as a procedural, accountability-driven expansion of government oversight: they foreground CAISI’s pre-deployment evaluations, link moves to Commerce directives and an interagency working group, highlight Anthropic’s flagged supply-chain risk alongside a "productive" meeting, and lean on anonymous sources to emphasize unfolding policy process.