The technology firms will voluntarily submit their latest AI models to the US Department of Commerce’s Center for AI Standards and Innovation (CAISI) for evaluation amid growing concerns around safety and national security.

The expanded agreements build on earlier arrangements made during the Joe Biden administration with companies including OpenAI and Anthropic, and will focus on testing AI capabilities, security risks and industry best practices.

According to BBC News, CAISI director Chris Fall said the collaborations would help the agency "scale our work in the public interest at a critical moment".

Google’s flagship AI system is Gemini, developed through its DeepMind subsidiary, while Microsoft’s best known platform is Copilot.

Elon Musk-owned xAI currently offers Grok, which has previously faced criticism over controversial image generation features.

On Tuesday (05.05.26), CAISI revealed it has already conducted 40 evaluations on AI systems, including unreleased “state-of-the-art models”, although it did not identify which tools had been examined or blocked from public launch.

Microsoft said in a corporate blog post that national security testing “must be a collaborative endeavour with governments”.