The US is taking AI under control: Google and Microsoft will show models before the release

The largest technological corporations Google, Microsoft and xAI reached a strategic agreement with the US government. As part of the new arrangements, they will provide the ministry with early access to their latest AI models to thoroughly test their capabilities and security before official release. This was reported by the US Department of Commerce through the Bloomberg agency.

The move brings together Google, Microsoft and Elon Musk’s xAI company with OpenAI and Anthropic, which already work with the Center for AI Standards and Innovation. This government agency, operating within the structure of the US Department of Commerce, acts as a central hub for the evaluation and analysis of AI systems before entering the wider market.

Features of verification processes and security metrics

The Ministry declares high activity in the field of audit. Starting from 2024, experts have already conducted more than 40 comprehensive assessments of artificial intelligence systems. It is important that this list also includes those models that were still presented to the public today. The main purpose of such checks is to prevent potential risks in the field of national security and information stability.

Political context

Intensification of government surveillance is taking place against the background of an active transformation of the White House’s high-tech policy. In July 2025, the AI ​​Action Plan strategic document was published by the Donald Trump administration. This plan envisages the creation of a full-fledged state ecosystem for evaluating models aimed at preventing threats to national security. Chris Fall, director of the Center for AI Standards, emphasized that the new partnerships will help scale the organization’s work for the benefit of society at a critical time for the industry.

Technical nuances

Increased attention from the state is largely driven by the development of complex architectures such as Anthropic’s Mythos. Last month, the developers confirmed that their system is capable of effectively detecting critical cyber security vulnerabilities. This fact has prompted US agencies and the White House to become more involved in regulating the field of generative AI development.