Category: AI / Big Tech / National Security


Washington just got a foot in the door of America's most powerful AI labs — before the rest of the world even knows what's behind it. Microsoft, Google, and Elon Musk's xAI have agreed to give the U.S. government early access to their newest AI models for national security testing, in a deal brokered through the Center for AI Standards and Innovation (CAISI) at the Department of Commerce.

The trigger? Washington is on edge over the hacking capabilities of Anthropic's newly unveiled Mythos model — a system powerful enough to send alarm bells ringing from Silicon Valley to the Pentagon. With frontier AI now capable of supercharging cyberattacks and military misuse, U.S. officials aren't waiting around for these tools to hit the open market before they take a long, hard look under the hood.

Under the agreement, CAISI will evaluate the models before deployment — assessing their capabilities, stress-testing for unexpected behaviors, and developing shared datasets and workflows with the companies. Notably, developers are handing over versions of their models with safety guardrails stripped back so the government can really put them through the wringer. CAISI says it has already knocked out more than 40 evaluations, including models that haven't yet seen the light of day publicly.

This isn't a Biden-era holdover. The agreement delivers on a pledge the Trump administration made in July 2025 to partner with tech companies to vet AI models for national security risks — and it builds squarely on earlier deals cut with OpenAI and Anthropic back in 2024. CAISI, which used to go by the U.S. Artificial Intelligence Safety Institute under Biden, has been repositioned as the government's go-to hub for frontier AI model testing.

The bigger picture here is hard to miss: frontier AI is no longer being treated like just another tech product that gets regulated after the fact. Washington is planting its flag at the pre-release stage — making clear that the most powerful AI systems in the world are now considered strategic national infrastructure. The era of "ship it and figure it out later" is running out of runway fast.

Official source:
U.S. Department of Commerce / CAISI / Reuters

Original official source URL:
https://www.commerce.gov/

Source article URL:
https://www.investing.com/news/stock-market-news/microsoft-xai-and-google-will-share-ai-models-with-us-govt-for-security-reviews-4658803