The US government has quietly scrubbed a major announcement detailing new CAISI AI vetting agreements to inspect unreleased AI models from Google, xAI, and Microsoft. The sudden disappearance of the web page, originally published by the Commerce Department’s Center for AI Standards and Innovation (CAISI) on May 5, 2026, has sparked questions about the stability of federal AI oversight.
The now-deleted agreements granted federal regulators pre-deployment access to evaluate frontier AI capabilities before they reach the general public. This initiative was designed to build upon similar safety testing partnerships established with Anthropic and OpenAI in 2024, aiming to advance the state of AI security and information-sharing.
Today, the Center for AI Standards and Innovation (CAISI) at the Department of Commerce’s National Institute of Standards and Technology announced new agreements with Google DeepMind, Microsoft and xAI. Through these expanded industry collaborations, CAISI will conduct pre-deployment evaluations and targeted research to better assess frontier AI capabilities and advance the state of AI security. These agreements build on previously announced partnerships, which have been renegotiated to reflect CAISI’s directives from the secretary of commerce and America’s AI Action Plan.
- Deleted CAISI Announcement
As first spotted by Reuters, the original URL initially returned a "Sorry, we cannot find that page" error before being altered to redirect visitors to the main CAISI homepage. Neither the White House nor the Commerce Department has provided an official explanation for the retraction, though archived versions of the page explicitly stated the deals were meant to ensure a clear understanding of "the state of international AI competition."
The Transparency Gap in Frontier AI Testing
The silent removal of this announcement highlights a growing friction point between federal regulators and Silicon Valley. The archived text explicitly framed these agreements around national security directives and international competition. Pulling the page suggests either a premature publication before final legal ink dried, or sudden pushback from the involved tech giants regarding the scope of government access to their proprietary models.
If CAISI intends to enforce rigorous pre-deployment evaluations on companies like Google and xAI, maintaining public transparency will be just as critical as the testing itself. Operating these vetting agreements in the dark risks undermining public trust in how the government handles the most powerful, unreleased AI systems in the world.
