Microsoft, Google, xAI testing deal removed from US gov’t website

The US Commerce Department has removed from its website details of a May 5 agreement involving Microsoft, Google, and xAI.

Under the agreement, the companies would give the US government early access to new AI models for security testing before public release.

The original announcement link showed an error on May 11 in Washington and later redirected to the Center for AI Standards and Innovation, the government unit handling the tests.

The deal came as US officials warned that advanced AI systems could pose national security risks including cyberattacks and military misuse, but the reason for deleting the page was not clear.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

This deal extends work across administrations to examine AI risks

The voluntary program builds on agreements the government made with OpenAI and Anthropic, an AI startup, in 2024 during the Biden administration 1.
The testing will be strict. The government’s Center for AI Standards and Innovation (CAISI), the unit running the tests, plans to assess the models in classified settings 2.
Reviews will examine AI models with safeguards scaled back or removed. That will help CAISI gauge their full capabilities, including national security risks 2.
The agreement suggests a shift in Trump administration policy. It comes as the administration backs an “AI Action Plan” meant to “remove red tape and onerous regulation” around the technology 3.

The deleted announcement adds to questions about US AI security reviews

Removing the announcement adds to uncertainty around the US approach to AI policy. Officials did not explain why the page came down 4.
President Trump has said too much regulation could slow innovation and give China an edge. His administration has also pushed to remove barriers to AI development 1.
National security concerns around powerful AI systems remain part of the debate, including risks such as cyberattacks and military misuse. Anthropic’s unreleased model “Mythos” has also surfaced in those discussions 4.
This uncertainty could make planning harder for AI companies. They face pressure to ship products fast while also dealing with government security reviews that may lack transparency.